DEVICE, METHOD, AND APP FOR FACILITATING SLEEP

Abstract

A device, system, and method for facilitating a sleep cycle in a subject, comprising selecting a waveform from a plurality of waveforms derived from brainwaves of at least one sleeping donor, wherein said waveform corresponds to at least one specific stage of sleep; and stimulating the subject with at least one stimulus, wherein said at least one stimulus is at least one of an auditory stimulus and a visual stimulus modulated with the selected waveform to entrain the brain of the subject with the selected waveform to facilitate sleep in the subject.

Description

CROSS REFERENCE TO RELATED APPLICATIONS

The present application is a non-provisional of, and claims benefit of priority under 35 U.S.C. § 119(e) from U.S. Provisional Patent Application No. 62/862,656, filed Jun. 17, 2019, and from U.S. Provisional Patent Application No. 62/852,877, filed May 24, 2019, each of which is expressly incorporated herein by reference in its entirety.

FIELD OF THE INVENTION

The present invention generally relates to the field of neuromodulation and neuro enhancement and more specifically to systems, methods and applications for improving achievement and/or maintenance of sleep.

BACKGROUND OF THE INVENTION

Each reference and document cited herein is expressly incorporated herein by reference in its entirety, for all purposes.

Brain Computer Interface (BCI): sometimes called a neural-control interface (NCI), mind-machine interface (MMI), direct neural interface (DNI), or brain-machine interface (BMI), is a communication pathway between a brain and an external computerized device. BCI may allows for bidirectional information flow. BCIs are often directed at researching, mapping, assisting, augmenting, or repairing human cognitive or sensory-motor functions. See, en.wikipedia.org/wiki/Brain-computer_interface. A bidirectional adaptive BCI controlling computer buyer by an anticipatory brain potential, the Contingent Negative Variation (CNV) potential has been reported. The experiment described how an expectation state of the brain, manifested by CNV, controls in a feedback loop the S2 buyer in the S1-S2-CNV paradigm. The obtained cognitive wave representing the expectation learning in the brain is named Electroexpectogram (EXG). Electroencephalography (EEG) is the most studied non-invasive interface, mainly due to its fine temporal resolution, ease of use, portability and low set-up cost. See Reference List Table 1.

Time in a biological manner. Almost everything in biology is subject to change overtime. These changes occur on many different time scales, which vary greatly. For example, there are evolutionary changes that affect entire populations over time rather than a single organism. Evolutionary changes are often slower than a human time scale that spans many years (usually a human lifetime). Faster variations of the timing and duration of biological activity in living organisms occur, for example, in many essential biological processes in everyday life: in humans and animals, these variations occur, for example, in eating, sleeping, mating, hibernating, migration, cellular regeneration, etc. Other fast changes may include the transmission of a neural signal, for example, through a synapse such as the calyx of held, a particularly large synapse in the auditory central nervous system of mammals that can reach transmission frequencies of up to 50 Hz. nth recruitment modulation, the effective frequencies can be higher. A single nerve impulse can reach a speed as high as one hundred meters (0.06 mile) per second (Kraus, David. Concepts in Modern Biology. New York: Globe Book Company, 1969:170). Myelination of axons can increase the speed of transmission by segmenting the membrane depolarization process.

Many of these changes over time are repetitive or rhythmic and are described as some frequency or oscillation. The field of chronobiology, for example, examines such periodic (cyclic) phenomena in living organisms and their adaptation, for example, to solar and lunar-related rhythms [DeCoursey et al. (2003).] These cycles are also known as biological rhythms. The related terms chronomics and chronome have been used in some cases to describe either the molecular mechanisms involved in chronobiological phenomena or the more quantitative aspects of chronobiology, particularly where comparison of cycles between organisms is required. Chronobiological studies include, but are not limited to, comparative anatomy, physiology, genetics, molecular biology, and behavior of organisms within biological rhythms mechanics [DeCoursey et al. (2003).]. Other aspects include epigenetics, development, reproduction, ecology, and evolution.

The most important rhythms in chronobiology are the circadian rhythms, roughly 24-hour cycles shown by physiological processes in all these organisms. It is regulated by circadian docks. The circadian rhythms can be further broken down into routine cycles during the 24-hour day [Nelson R J. 2005. An Introduction to Behavioral Endocrinology. Sinauer Associates, Inc.: Massachusetts. Pg. 587.] All animals can be classified according to their activity cycles: Diurnal, which describes organisms active during daytime; Nocturnal, which describes organisms active in the night and Crepuscular, which describes animals primarily active during the dawn and dusk hours (ex: white-tailed deer, some bats).

While circadian rhythms are defined as regulated by endogenous processes, other biological cycles may be regulated by exogenous signals. In some cases, multi-trophic systems may exhibit rhythms driven by the circadian dock of one of the members (which may also be influenced or reset by external factors).

Many other important cycles are also studied, including Infradian rhythms, which are cycles longer than a day. Examples include circannual or annual cycles that govern migration or reproduction cycles in many plants and animals, or the human menstrual cycle; Ultradian rhythms, which are cycles shorter than 24 hours, such as the 90-minute REM cycle, the 4-hour nasal cycle, or the 3-hour cycle of growth hormone production; Tidal rhythms, commonly observed in marine life, which follow the roughly 12.4-hour transition from high to low tide and back; Lunar rhythms, which follow the lunar month (29.5 days). They are relevant, for example, to marine life, as the level of the tides is modulated across the lunar cycle; and Gene oscillations—some genes are expressed more during certain hours of the day than during other hours.

Within each cycle, the time period during which the process is more active is called the acrophase [Refinetti, Roberto (2006). Circadian Physiology. CRC Press/Taylor & Francis Group. ISBN 0-8493-2233-2. Lay summary]. When the process is less active, the cycle is in its bathyphase or trough phase. The particular moment of highest activity is the peak or maximum; the lowest point is the nadir. How high (or low) the process gets is measured by the amplitude.

The sleep cycle and the ultradian rhythms: The normal cycle of sleep and wakefulness implies that, at specific times, various neural systems are being activated while others are being turned off. A key to the neurobiology of sleep is, therefore, to understand the various stages of sleep. In 1953, Nathaniel Kleitman and Eugene Aserinksy showed, using electroencephalographic (EEG) recordings from normal human subjects, that sleep comprises different stages that occur in a characteristic sequence.

Humans descend into sleep in stages that succeed each other over the first hour or so after retiring. These characteristic stages are defined primarily by electroencephalographic criteria. Initially, during “drowsiness,” the frequency spectrum of the electroencephalogram (EEG) is shifted toward lower values, and the amplitude of the cortical waves slightly increases. This drowsy period, called stage I sleep, eventually gives way to light or stage II sleep, which is characterized by a further decrease in the frequency of the EEG waves and an increase in their amplitude, together with intermittent high-frequency spike dusters called sleep spindles. Sleep spindles are periodic bursts of activity at about 10-12 Hz that generally last 1 or 2 seconds and arise as a result of interactions between thalamic and cortical neurons. In stage III sleep, which represents moderate to deep sleep, the number of spindles decreases, whereas the amplitude of low-frequency waves increases still more. In the deepest level of sleep, stage N sleep, the predominant EEG activity consists of low-frequency (1-4 Hz), high-amplitude fluctuations called delta waves, the characteristic slow waves for which this phase of sleep is named. The entire sequence from drowsiness to deep stage N sleep usually takes about an hour.

These four sleep stages are called non-rapid eye movement (non-REM or NREM) sleep, and its most prominent feature is the slow-wave (stage IV) sleep. Sometimes, stages III and IV are combined and referred to jointly as the stage III sleep. It is most difficult to awaken people from slow-wave sleep; hence, it is considered to be the deepest stage of sleep. Following a period of slow-wave sleep, however, EEG recordings show that the stages of sleep reverse to reach a quite different state called rapid eye movement, or REM, sleep. In REM sleep, the EEG recordings are remarkably similar to that of the awake state. This mode is bizarre: a dreamer's brain becomes highly active while the body's muscles are paralyzed, and breathing and heart rate become erratic. After about 10 minutes in REM sleep, the brain typically cycles back through the non-REM sleep stages. Slow-wave sleep usually occurs again in the second period of this continual cycling, but not during the rest of the night. On average, four additional periods of REM sleep occur, each having longer than the preceding cycle durations.

The sleep cycle is an oscillation between the non-REM (including slow-waves) and REM phases of sleep. It is sometimes called the ultradian sleep cycle, sleep-dream cycle, or REM-NREM cycle, to distinguish it from the circadian alternation between sleep and wakefulness. In humans, this cycle takes on average between land 2 hours (approximately 90 min).

The timing of sleep cycles can be observed on EEG by marked distinction in brainwaves manifested during REM and non-REM sleep. Delta wave activity, correlating with slow-wave (deep) sleep, in particular, shows regular oscillations throughout a night's sleep. Secretions of various hormones, including renin, growth hormone, and prolactin, correlate positively with delta-wave activity, whereas secretion of thyroid-stimulating hormone correlates inversely. Heart rate variability, well-known to increase during REM, also correlates inversely with delta-wave oscillations over the 90-minute cycle.

Homeostatic functions, especially thermoregulation, normally occur during non-REM sleep, but not during REM sleep. During REM sleep, body temperature tends to drift from its mean level, and during non-REM sleep, to return to normal. The alternation between the stages, therefore, maintains body temperature within an acceptable range.

In humans, the transition between non-REM and REM is abrupt; in other animals, less so.

Different models have been proposed to elucidate the complex rhythm of electrochemical processes that result in the regular alternation of REM and non-REM sleep. Monoamines are active during non-REM stages but not during REM stages, whereas acetylcholine is more active during REM sleep. The reciprocal interaction model proposed in the 1970s suggested a cyclic give and take between these two systems. More recent theories such as the “flip-flop” model proposed in the 2000s include the regulatory role of in inhibitory neurotransmitter gamma-aminobutyric acid (GABA).

The average length of the sleep cycle in an adult man is 90 minutes. N1 (NREM stage 1) is when the person is drowsy or awake to falling asleep. Brain waves and muscle activity start to decrease at this stage. N2 is when the person experiences a light sleep. Eye movement has stopped by this time. Brain wave frequency and muscle tonus is decreased. The heart rate and body temperature go down. N3 or even N4 is the most difficult stages to be awakened. Every part of the body is now relaxed, breathing is slowed, blood pressure and body temperature are reduced. REM sleep is a unique state, in which dreams usually occur. The brain is awake, and body paralyzed. This unique stage is usually when the person is in the deepest stage of sleep and dreams. The average length of a sleep cycle usually thought of as 90 min. Some sources give it 90-110 minutes or an even wider range of 80-120 minutes. A seven-eight-hour sleep usually includes five cycles, the middle two of which tend to be longer. REM takes up more of the cycle as the night goes on.

When falling asleep, a series of highly orchestrated events puts the brain to sleep in the above-mentioned stages. Technically, sleep starts in the brain areas that produce slow-wave sleep (SWS). It has been shown that two groups of cells—the ventrolateral preoptic nucleus in the hypothalamus and the parafacial zone in the brain stem—are involved in prompting SWS. When these cells are activated, it triggers a loss of consciousness. After SWS, REM sleep begins. The purpose of REM sleep remains a biological mystery, despite our growing understanding of its biochemistry and neurobiology. It has been shown that a small group of cells in the brain stem, called the subcoeruleus nucleus, control REM sleep. When these cells become injured or diseased, people do not experience the muscle paralysis associated with REM sleep, which can lead to REM sleep behavior disordera serious condition in which the afflicted violently act out their dreams. For reasons that are not clear, the amount of REM sleep each day decreases from about 8 hours at birth to 2 hours at 20 years, to only about 45 minutes at 70 years of age. See Mallick, B. N.; S. R. Pand-Perumal; RobedW, McCarley; and Adrian R. Morrison (2011). Rapid Eye Movement Sleep: Regulation and Function. Cambridge University Press. ISBN 978-0-521-11680-0; Nir, and Tononi, “Dreaming and the Brain: from Phenomenology to Neurophysiology.” Trends in Cognitive Sciences, vol. 14, no. 2, 2010, pp. 88-100; and Varela, F., Engel, J., Wallace, B., & Thupten, lip. (1997). Sleeping, dreaming, and dying: An exploration of consciousness with the Dalai Lama.

Mental State: A mental state is a state of mind thata subject is in. Some mental states are pure and unambiguous, while humans are capable of complex states that are a combination of mental representations, which may have in their pure state contradictory characteristics. There are several paradigmatic states of mind thata subject has: love, hate, pleasure, fear, and pain. Mental states can also include a waking state, a sleeping state, a flow (or being in the “zone”), and a mood (a mental state). A mental state is a hypothetical state that corresponds to thinking and feeling and consists of a conglomeration of mental representations. A mental state is related to an emotion, though it can also relate to cognitive processes. Because the mental state itself is complex and potentially possess inconsistent attributes, clear interpretation of mental state through external analysis (other than self-reporting) is difficult or impossible. However, some studies repod that certain attributes of mental state or thought processes may, in fact, be determined through passive monitoring, such as EEG, or fMRI with some degree of statistical reliability. In most studies, the characterization of mental state was an endpoint, and the raw signals, after statistical classification or semantic labeling, are superseded. The remaining signal energy treated as noise. Current technology does not permit a precise abstract encoding or characterization of the full range of mental states based on neural correlates of mental state.

Brain: The brain is a key part of the central nervous system, enclosed in the skull. In humans, and mammals more generally, the brain controls both autonomic processes, as well as cognitive processes. The brain (and to a lesser extent, the spinal cord) controls all volitional functions of the body and interprets information from the outside world. Intelligence, memory, emotions, speech, thoughts, movements, and creativity are controlled by the brain. The central nervous system also controls autonomic functions and many homeostatic and reflex actions, such as breathing, heart rate, etc. The human brain consists of the cerebrum, cerebellum, and brainstem. The brainstem includes the midbrain, the pons, and the medulla oblongata. Sometimes the diencephalon, the caudal part of the forebrain, is included.

The brainstem has many basic functions, including heart rate, breathing, sleeping, and eating. The skull imposes a barrier to electrical access to the brain functions, and in a healthy human, breaching the dura to access the brain is highly disfavored. The result is that electrical readings of brain activity are filtered by the dura, the cerebrospinal fluid, the skull, the scalp, skin appendages (e.g., hair), resulting in a loss of potential spatial resolution and amplitude of signals emanating from the brain. While magnetic fields resulting from brain electrical activity are accessible, the spatial resolution using feasible sensors is also limited.

Technological advances now allow for the non-invasive recording of large quantities of information from the brain at multiple spatial and temporal scales. Examples include electroencephalogram (“EEG”) data using mufti-channel electrode arrays placed on the scalp or inside the brain, magnetoencephalography (“MEG”), magnetic resonance imaging (“MRI”), functional data using functional magnetic resonance imaging (“fMRI”), positron emission tomography (“PET”), near-infrared spectroscopy (“NIRS”), single-photon emission computed tomography (“SPECT”), and others. Noninvasive neuromodulation technologies have also been developed that can modulate the pattern of neural activity, and thereby cause altered behavior, cognitive states, perception, and motor output. Integration of noninvasive measurement and neuromodulation techniques for identifying and transplanting brain states from neural activity would be very valuable for clinical therapies, such as brain stimulation and related technologies often attempting to treat disorders of cognition. See, Mehmetali Gülpinar, Berrak C Yeğen, “The Physiology of Learning and Memory: Role of Peptides and Stress”, Current Protein and Peptide Science, 2004(5); www.researchgate.net/publication/8147320_The_Physiology_of_Learning_and_Memory_Role_of_Peptides_and_Stress. Deep brain stimulation is described in NIH Research Matters, “A noninvasive deep brain stimulation technique”, (2017); Brainworks, “QEEG Brain Mapping”; and Carmon, A., Mor, J., & Goldberg, J. (1976). Evoked cerebral responses to noxious thermal stimuli in humans. Experimental Brain Research, 25(1), 103-107.

Mental State: A number of studies repod that certain attributes of mental state or thought processes may, in fact, be determined through passive monitoring, such as EEG, with some degree of statistical reliability. In most studies, the characterization of mental state was an endpoint, and the raw signals, after statistical classification or semantic labeling, are superseded and the remaining signal energy treated as noise.

Neural Correlates: A neural correlate of a sleep state is an electro-neuro-biological state or the state assumed by some biophysical subsystem of the brain, whose presence necessarily and regularly correlates with such specific sleep states. All properties credited to the mind, including consciousness, emotion, and desires are thought to have direct neural correlates. Neural correlates of a sleep state can be defined as the minimal set of neuronal oscillations that correspond to the given sleep stage.

Brainwaves: At the root of all our thoughts, emotions, and behaviors is the communication between neurons within our brains, a rhythmic or repetitive neural activity in the central nervous system. The oscillation can be produced by a single neuron or by synchronized electrical pulses from ensembles of neurons communicating with each other. The interaction between neurons can give rise to oscillations at a different frequency than the firing frequency of individual neurons. The synchronized activity of large numbers of neurons produces macroscopic oscillations, which can be observed in an electroencephalogram. They are divided into bandwidths to describe their purported functions or functional relationships. Oscillatory activity in the brain is widely observed at different levels of organization and is thought to play a key role in processing neural information. Numerous experimental studies support a functional role of neural oscillations. A unified interpretation, however, is still not determined. Neural oscillations and synchronization have been linked to many cognitive functions such as information transfer, perception, motor control, and memory. Electroencephalographic (EEG) signals are relatively easy and safe to acquire, have a long history of analysis, and can have high dimensionality, e.g., up to 128 or 256 separate recording electrodes. While the information represented in each electrode is not independent of the others, and the noise in the signals high, there is much information available through such signals that has not been fully characterized to date.

Brainwaves have been widely studied in neural activity generated by large groups of neurons, mostly by EEG. In general, EEG signals reveal oscillatory activity (groups of neurons periodically firing in synchrony), in specific frequency bands: alpha (7.5-125 Hz) that can be detected from the occipital lobe during relaxed wakefulness and which increases when the eyes are closed; delta (1-4 Hz), theta (4-8 Hz), beta (13-30 Hz), low gamma (30-70 Hz), and high gamma (70-150 Hz) frequency bands, where faster rhythms such as gamma activity have been linked to cognitive processing. Higher frequencies imply multiple groups of neurons firing in coordination, either in parallel or in series, or both, since individual neurons do not fire at rates of 100 Hz. Neural oscillations of specific characteristics have been linked to cognitive states, such as awareness and consciousness and different sleep stages. See, Chang-Hwan Im, Computational EEG Analysis: Methods and Applications (Biological and Medical Physics, Biomedical Eng.), Sep. 11, 2019.

Nyquist Theorem states that the highest frequency that can be accurately represented is one-haft of the sampling rate. Practically, the sampling rate should be ten times higher than the highest frequency of the signal. (See, www.slideshare.net/ertyk/eeg-examples). While EEG signals are largely band limited, the superimposed noise may not be. Further, the EEG signals themselves represent components from a large number of neurons, which fire independently. Therefore, large bandwidth signal acquisition may have utility.

It is a useful analogy to think of brainwaves as musical notes. Like in a symphony, the higher and lower frequencies link and cohere with each other through harmonics, especially when one considers that neurons may be coordinated not only based on transitions, but also on phase delay. Oscillatory activity is observed throughout the central nervous system at all levels of organization. The dominant neuro oscillation frequency is associated with a respective mental state. The functions of brainwaves are wide-ranging and vary for different types of oscillatory activity. Neural oscillations also play an important role in many neurological disorders.

In standard EEG recording practice, 19 recording electrodes are placed uniformly on the scalp (the International 10-20 System). In addition, one or two reference electrodes (often placed on earlobes) and a ground electrode (often placed on the nose to provide amplifiers with reference voltages) are required. However, additional electrodes may add minimal useful information unless supplemented by computer algorithms to reduce raw EEG data to a manageable form. When large numbers of electrodes are employed, the potential at each location may be measured with respect to the average of all potentials (the common average reference), which often provides a good estimate of potential at infinity. The common average reference is not appropriate when electrode coverage is sparse (perhaps less than 64 electrodes). See, Paul L Nunez and Ramesh Srinivasan (2007) Electroencephalogram. Scholarpedia, 2(2):1348, scholarpedia.org/article/Electroencephalogram. Dipole localization algorithms may be useful to determine spatial emission patterns in EEG.

Scalp potential may be expressed as a volume integral of dipole moment per unit volume over the entire brain provided P(r,t) is defined generally rather than in columnar terms. For the important case of dominant cortical sources, scalp potential may be approximated by the following integral over the cortical volume Θ,VS(r,t)=∫∫∫ΘG(r,r′)·P(r′,t)dΘ(r′). If the volume element dΘ(r′) is defined in terms of cortical columns, the volume integral may be reduced to an integral over the folded cortical surface. The time-dependence of scalp potential is the weighted sum of all dipole time variations in the brain, although deep dipole volumes typically make negligible contributions. The vector Green's function G(r,r′) contains all geometric and conductive information about the head volume conductor and weights the integral accordingly. Thus, each scalar component of the Green's function is essentially an inverse electrical distance between each source component and scalp location. For the idealized case of sources in an infinite medium of constant conductivity, the electrical distance equals the geometric distance. The Green's function accounts for the tissue's finite spatial extent and its in homogeneity and anisotropy. The forward problem in EEG consists of choosing a head model to provide G(r,r′) and carrying out the integral for some assumed source distribution. The inverse problem consists of using the recorded scalp potential distribution VS(r,t) plus some constraints (usual assumptions) on P(r,t) to find the best fit source distribution P(r,t). Since the inverse problem has no unique solution, any inverse solution depends critically on the chosen constraints, for example, only one or two isolated sources, distributed sources confined to the cortex, or spatial and temporal smoothness criteria. High-resolution EEG uses the experimental scalp potential VS(r,t) to predict the potential on the dura surface (the unfolded membrane surrounding the cerebral cortex) VD(r,t). This may be accomplished using a head model Green's function G(r,r′) or by estimating the surface Laplacian with either spherical or 3D splines. These two approaches typically provide very similar dura potentials VD(r,t); the estimates of dura potential distribution are unique subject to head model, electrode density, and noise issues.

In an EEG recording system, each electrode is connected to one input of a differential amplifier (one amplifier per pair of electrodes); a common system reference electrode (or synthesized reference) is connected to the other input of each differential amplifier. These amplifiers amplify the voltage between the active electrode and the reference (typically 1,000-100,000 times, or 60-100 dB of voltage gain). The amplified signal is digitized via an analog-to-digital converter, after being passed through an anti-aliasing filter. Analog-to-digital sampling typically occurs at 256-512 Hz in clinical scalp EEG; sampling rates of up to 20 kHz are used in some research applications. The EEG signals can be captured with open source hardware such as OpenBCI, and the signal can be processed by freely available EEG software such as EEGLAB or the Neurophysiological Biomarker Toolbox. A typical adult human EEG signal is about 10 μV to 100 μV (scalp) and about 10-20 mV (subdural electrodes).

Delta wave (en.wikipedia.org/wiki/Delta_wave) is the frequency range up to 4 Hz. It tends to be the highest in amplitude and the slowest waves. It is normally seen in adults in NREM (en.wikipedia.org/wiki/NREM). It is also seen normally in babies. It may occur focally with subcortical lesions and in general distribution with diffuse lesions, metabolic encephalopathy hydrocephalus or deep midline lesions. It is, usually, most prominent frontally in adults (e.g., FIRDA-frontal intermittent rhythmic delta) and posteriorly in children (e.g., OIRDA-occipital intermittent rhythmic delta).

Theta is the frequency range from 4 Hz to 7 Hz. Theta is normally seen in young children. It may be seen in drowsiness or arousal in older children and adults; it can also be seen in meditation. Excess theta for age represents abnormal activity. It can be seen as a focal disturbance in focal subcortical lesions; it can be seen in the generalized distribution in diffuse disorder or metabolic encephalopathy or deep midline disorders or some instances of hydrocephalus. On the contrary, this range has been associated with reports of relaxed, meditative, and creative states.

Alpha is the frequency range from 7 Hz to 14 Hz. This was the “posterior basic rhythm” (also called the “posterior dominant rhythm” or the “posterior alpha rhythm”), seen in the posterior regions of the head on both sides, higher in amplitude on the dominant side. It emerges with the closing of the eyes and with relaxation and attenuates with eye opening or mental exertion. The posterior basic rhythm is slower than 8 Hz in young children (therefore technically in the theta range). In addition to the posterior basic rhythm, there are other normal alpha rhythms such as the sensorimotor, or mu rhythm (alpha activity in the contralateral sensory and motor cortical areas) that emerges when the hands and arms are idle; and the “third rhythm” (alpha activity in the temporal or frontal lobes). Alpha can be abnormal; for example, an EEG that has diffuse alpha occurring in a coma and is not responsive to external stimuli is referred to as “alpha coma.”

Beta is the frequency range from 15 Hz to about 30 Hz. It is usually seen on both sides in symmetrical distribution and is most evident frontally. Beta activity is closely linked to motor behavior and is generally attenuated during active movements. Low-amplitude beta with multiple and varying frequencies is often associated with active, busy, or anxious thinking and active concentration. Rhythmic beta with a dominant set of frequencies is associated with various pathologies, such as Dup15q syndrome, and drug effects, especially benzodiazepines. It may be absent or reduced in areas of cortical damage. It is the dominant rhythm in patients who are alert or anxious or who have their eyes open.

Gamma is the frequency range of approximately 30-100 Hz. Gamma rhythms are thought to represent binding of different populations of neurons together into a network to carry out a certain cognitive or motor function.

Mu range is 8-13 Hz and partly overlaps with other frequencies. It reflects the synchronous firing of motor neurons in a rest state. Mu suppression is thought to reflect motor mirror neuron systems because when an action is observed, the pattern extinguishes, possibly because of the normal neuronal system and the mirror neuron system “go out of sync” and interfere with each other. (en.wikipedia.org/wiki/Electroencephalography). See Reference List Table 2.

All sleep stages are associated with frequencies below 13 Hz-delta (1-4 Hz), theta (4-8 Hz), and alpha (8-12 Hz). While these frequencies may be reproduced in transcranial electric (or magnetic) stimulation, or via sensory stimulation with light, any attempts to reproduce these frequencies for stimulation via sound ran into problems associated with infrasound, defined as any sound below 20 Hz frequency. Firstly, it is difficult to generate infrasound through acoustic speakers. Earbuds are too small for that and so are most regular speakers. Specialized large subwoofers with circular design or sound guides may be used, but tend to be impractical.

Binaural beats: (See en.wikipedia.org/wiki/Beat_(acoustics)#Binaural_beats) A binaural beat is an auditory illusion perceived when two different pure-tone sine waves, both with frequencies lower than 1500 Hz, with less than a 40 Hz difference between them, are presented to a listener dichotically (one through each ear). A more serious problem is the effect of the infrasound on human health. While many animals (e.g., elephants and wales) communicate via infrasound, in humans, infrasound causes undesirable effects including send of panic, fear, and anxiety. Prolonged exposure to infrasound could be dangerous to human health. See, for example, Persinger, M. A. Nat Hazards (2014) 70: 501.doi.org/10.1007/s11069-013-0827-3. These problems are addressed using binaural beats. For example, if a 530 Hz pure tone is presented to a subject's right ear, while a 520 Hz pure tone is presented to the subject's left ear, the listener will perceive the auditory illusion of a third tone, in addition to the two pure-tones presented to each ear. The third sound is called a binaural beat, and in this example would have a perceived pitch correlating to a frequency of 10 Hz, that being the difference between the 530 Hz and 520 Hz pure tones presented to each ear.

Binaural-beat perception originates in the inferior colliculus of the midbrain and the superior olivary complex of the brainstem, where auditory signals from each ear are integrated and precipitate electrical impulses along neural pathways through the reticular formation up the midbrain to the thalamus, auditory cortex, and other cortical regions. Binaural beats are widely used in brain stimulation.

EEG AND qEEG: An EEG electrode will mainly detect the neuronal activity in the brain region just beneath it. However, the electrodes receive the activity from thousands of neurons. One square millimeter of cortex surface, for example, has more than 100,000 neurons. It is only when the input to a region is synchronized with electrical activity occurring at the same time that simple periodic waveforms in the EEG become distinguishable. The temporal pattern associated with specific brainwaves can be digitized and encoded a non-transient memory, and embodied in or referenced by, computer software.

EEG (electroencephalography) and MEG (magnetoencephalography) are available technologies to monitor brain electrical activity. Each generally has sufficient temporal resolution to follow dynamic changes in brain electrical activity. Electroencephalography (EEG) and quantitative electroencephalography (qEEG) are electrophysiological monitoring methods that analyze the electrical activity of the brain to measure and display patterns that correspond to cognitive states and/or diagnostic information. It is typically noninvasive, with the electrodes placed on the scalp, although invasive electrodes are also used in some cases. EEG signals may be captured and analyzed by a mobile device, often referred to as “brain wearables.” There are a variety of “brain wearables” readily available on the market today. EEGs can be obtained with a non-invasive method where the aggregate oscillations of brain electric potentials are recorded with numerous electrodes attached to the scalp of a person. Most EEG signals originate in the brain's outer layer (the cerebral cortex), believed largely responsible for our thoughts, emotions, and behavior. Cortical synaptic action generates electrical signals that change in the 10 to 100-millisecond range. Transcutaneous EEG signals are limited by the relatively insulating nature of the skull surrounding the brain, the conductivity of the cerebrospinal fluid and brain tissue, relatively low amplitude of individual cellular electrical activity, and distances between the cellular current flows and the electrodes. EEG is characterized by: (1) Voltage; (2) Frequency; (3) Spatial location; (4) Inter-hemispheric symmetries; (5) Reactivity (reaction to state change); (6) Character of waveform occurrence (random, serial, continuous); and (7) Morphology of transient events. EEGs can be separated into two main categories. Spontaneous EEG which occur in the absence of specific sensory stimuli and evoked potentials (EPs) which are associated with sensory stimuli like repeated light flashes, auditory tones, finger pressure, or mild electric shocks. The latter is recorded, for example, by time averaging to remove effects of spontaneous EEG. Non-sensory triggered potentials are also known. EP's typically are time synchronized with the trigger, and thus have an organization principle. Event-related potentials (ERPs) provide evidence of a direct link between cognitive events and brain electrical activity in a wide range of cognitive paradigms. It has generally been held that an ERP is the result of a set of discrete stimulus-evoked brain events. Event-related potentials (ERPs) are recorded in the same way as EPs, but occur at longer latencies from the stimuli and are more associated with an endogenous brain state.

Typically, a magnetic sensor with sufficient sensitivity to individual cell depolarization or small groups is a superconducting quantum interference device (SQUID), which requires cryogenic temperature operation, either at liquid nitrogen temperatures (high-temperature superconductors, HIS) or liquid helium temperatures (low-temperature superconductors, LTS). However, current research shows the possible feasibility of room temperature superconductors (20 C). Magnetic sensing has an advantage, due to the dipole nature of sources, of having better potential volumetric localization; however, due to this added information, the complexity of signal analysis is increased.

In general, the electromagnetic signals detected represent action potentials, an automatic response of a nerve cell to depolarization beyond a threshold, which briefly opens conduction channels. The cells have ion pumps which seek to maintain a depolarized state. Once triggered, the action potential propagates along the membrane in two-dimensions, causing a brief high level of depolarizing ion flow. There is a quiescent period after depolarization that generally prevents oscillation within a single cell. Since the exon extends from the body of the neuron, the action potential will typically proceed along the length of the axon, which terminates in a synapse with another cell. While direct electrical connections between cells occur, often the axon releases a neurotransmitter compound into the synapse, which causes depolarization or hyperpolarization of the target cell. Indeed, the result may also be the release of a hormone or peptide, which may have a local or more distant effect.

The electrical fields detectable externally tend to not include signals which low-frequency signals, such as static levels of polarization, or cumulative depolarizing or hyperpolarizing effects between action potentials. In myelinated tracts, the current flows at the segments tend to be small, and therefore, the signals from individual cells are small. Therefore, the largest signal components are from the synapses and cell bodies. In the cerebrum and cerebellum, these structures are mainly in the cortex, which is largely near the skull, making electroencephalography useful, since it provides spatial discrimination based on electrode location. However, deep signals are attenuated and poorly localized. Magnetoencephalography detects dipoles, which derive from current flow, rather than voltage changes. In the case of a radially or spherically symmetric current flow within a short distance, the dipoles will tend to cancel, while net current flows long axons will reinforce. Therefore, an electroencephalogram reads a different signal than a magnetoencephalogram.

EEG-based studies of emotional specificity at the single-electrode level demonstrated that asymmetric activity at the frontal site, especially in the alpha (8-12 Hz) band, is associated with emotion. Voluntary facial expressions of smiles of enjoyment produce higher left frontal activation. Decreased left frontal activity is observed during the voluntary facial expressions of fear. In addition to alpha-band activity, theta band power at the frontal midline (Fm) has also been found to relate to emotional states. Pleasant (as opposed to unpleasant) emotions are associated with an increase in frontal midline theta power. Many studies have sought to utilize pattern classification, such as neural networks, statistical classifiers, clustering algorithms, etc., to differentiate between various emotional states reflected in EEG. Ekman and Davidson found that voluntary facial expressions of smiles of enjoyment produced higher left frontal activation (Ekman P, Davidson R J (1993) Voluntary Smiling Changes Regional Brain Activity. Psychol Sci 4: 342-345). Another study by Coan et al. found decreased left frontal activity during the voluntary facial expressions of fear (Coan J A, Allen J J, Harmon-Jones E (2001) Voluntary facial expression and hemispheric asymmetry over the frontal cortex. Psychophysiology 38: 912-925). Sammler and colleagues, for example, showed that pleasant (as opposed to unpleasant) emotion is associated with an increase in frontal midline theta power (Sammler D, Grigutsch M, Fritz T, Koelsch 5 (2007) Music and emotion: Electrophysiological correlates of the processing of pleasant and unpleasant music. Psychophysiology 44: 293-304). To further demonstrate whether these emotion-specific EEG characteristics are strong enough to differentiate between various emotional states, some studies have utilized a pattern classification analysis approach.

When filtering brainwave signals, it is often useful extract features from the noisy signal, where the features may be identified and correlated signal components can also be removed without distortion of the remaining components. While this can be directly employed when applied to a feature of interest, this may also be employed to simplify the brainwave signal for later analysis by removing characteristic features of a distinct patters from those sought. Thus, consideration of non-sleep states is useful.

Detecting different emotional states by EEG may be more appropriate using EEG-based functional connectivity. There are various ways to estimate EEG-based functional brain connectivity: correlation, coherence, and phase synchronization indices between each pair of EEG electrodes had been used. The assumption is theta higher correlation map indicates a stronger relationship between two signals. (Brazier M A Casby J U (1952) Cross-correlation and autocorrelation studies of electroencephalographic potentials. Electroen din neuro 4: 201-211). Coherence gives information similar to correlation but also includes the covariation between two signals as a function of frequency. (Cantero J L, Atienza M, Salas R M, Gomez C M (1999) Alpha EEG coherence in different brain states: an electrophysiological index of the arousal level in human subjects. Neurosci lett 271:167-70.) The assumption is that higher correlation indicates a stronger relationship between two signals. (Guevara M A, Corsi-Cabrera M (1996) EEG coherence or EEG correlation? Int J Psychophysiology 23:145-153; Cantero J L, Atienza M, Salas R M, Gomez C M (1999) Alpha EEG coherence in different brain states: an electrophysiological index of the arousal level in human subjects. Neurosci lett 271:167-70; Adler G, Brassen S, Jajcevic A (2003) EEG coherence in Alzheimer's dementia. J Neural Transm 110: 1051-1058; Deeny S P, Hillman C H, Janelle C M, Hatfield B D (2003) Cortico-cortical communication and superior performance in skilled marksmen: An EEG coherence analysis. J Sport Exercise Psy 25:188-204.) Phase synchronization among the neuronal groups estimated based on the phase difference between two signals is another way to estimate the EEG-based functional connectivity among brain areas. It is. (Franaszczuk P J, Bergey G K (1999) An autoregressive method for the measurement of synchronization of interictal and ictal EEG signals. Biol Cybern 81:3-9.)

A number of groups have examined emotional specificity using EEG-based functional brain connectivity. For example, Shin and Park showed that when emotional states become more negative at high room temperatures, correlation coefficients between the channels in temporal and occipital sites increase (Shin)-H, Park D-H. (2011) Analysis for Characteristics of Electroencephalogram (EEG) and Influence of Environmental Factors According to Emotional Changes. In Lee G, Howard D, Slczak D, editors. Convergence and Hybrid Information Technology. Springer Berlin Heidelberg, 488-500.) Hinrichs and Machleidt demonstrated that coherence decreases in the alpha band during sadness, compared to happiness (Hinrichs H, Machleidt W (1992) Basic emotions reflected in EEG-coherences. Intl Psychophysiol 13: 225-232). Miskovic and Schmidt found that EEG coherence between the prefrontal cortex and the posterior cortex increased while viewing highly emotionally arousing (i.e., threatening) images, compared to viewing neutral images (Miskovic V, Schmidt L A (2010) Cross-regional cortical synchronization during affective image viewing. Brain Res 1362:102-111). Costa and colleagues applied the synchronization index to detect interaction in different brain sites under different emotional states (Costa T, Rognoni E, Galati D (2006) EEG phase synchronization during emotional response to positive and negative film stimuli. Neurosci Lett 406:159-164). Costa's results showed an overall increase in the synchronization index among frontal channels during emotional stimulation, particularly during negative emotion (i.e., sadness). Furthermore, phase synchronization patterns were found to differ between positive and negative emotions. Costa also found that sadness was more synchronized than happiness at each frequency band and was associated with a wider synchronization both between the right and left frontal sites and within the left hemisphere. In contrast, happiness was associated with a wider synchronization between the frontal and occipital sites.

Different connectivity indices are sensitive to different characteristics of EEG signals. Correlation is sensitive to phase and polarity but is independent of amplitudes. Changes in both amplitude and phase lead to a change in coherence (Guevara M A, Corsi-Cabrera M (1996) EEG coherence or EEG correlation? Intl Psychophysiol 23:145-153). The phase synchronization index is only sensitive to a change in phase (Lachaux J P, Rodriguez E, Martinerie J, Varela H (1999) Measuring phase synchrony in brain signals. Hum Brain Mapp 8:194-208). A number of studies have tried to classify emotional states by means of recording and statistically analyzing EEG signals from the central nervous systems. See, for example: Lin Y P, Wang C H, Jung T P, Wu T L, leng S K, et al. (2010) EEG-Based Emotion Recognition in Music Listening. IEEE T Bio Med Eng 57:1798-1806; Murugappan M, Nagarajan R, Yaacob S (2010) Classification of human emotion from EEG using discrete wavelet transform. J Biomed Sci Eng 3: 390-396; Murugappan M, Nagarajan R, Yaacob S (2011) Combining Spatial Filtering and Wavelet Transform for Classifying Human Emotions Using EEG Signals. J Med. Bio. Eng. 31:45-51; Berkman E, Wong D K Guimaraes M P, Uy E T, Gross J J, et al. (2004) Brain wave recognition of emotions in EEG. Psychophysiology 41: S71-571; Chanel G, Kronegg J, Grandjean D, Pun T (2006) Emotion assessment: Arousal evaluation using EEG's and peripheral physiological signals. Multimedia Content Representation, Classification and Security 4105: 530-537; Hagiwara KIaM (2003) A Feeling Estimation System Using a Simple Electroencephalograph. IEEE International Conference on Systems, Man and Cybernetics. 4204-4209; and You-Yun Lee and Shulan Hsieh studied different emotional states by means of EEG-based functional connectivity patterns. They used emotional film dips to elicit three different emotional states.

The dimensional theory of emotion, which asserts that there are neutral, positive, and negative emotional states, may be used to classify emotional states because numerous studies have suggested that the responses of the central nervous system correlate with emotional valence and arousal. (See, for example, Davidson R J (1993) Cerebral Asymmetry and Emotion—Conceptual and Methodological Conundrums. Cognition Emotion 7: 115-138; Jones N A, Fox N A (1992) Electroencephalogram asymmetry during emotionally evocative films and its relation to positive and negative affectivity. Brain Cogn 20: 280-299; Schmidt L A, Trainor U (2001) Frontal brain electrical activity (EEG) distinguishes valence and intensity of musical emotions. Cognition Emotion 15: 487-500; Tomarken A I, Davidson R J, Henriques J B (1990) Resting frontal brain asymmetry predicts affective responses to films. J Pers Soc Psycho) 59:791-801.) As suggested by Mauss and Robins (2009), “measures of emotional responding appear to be structured along dimensions (e.g., valence, arousal) rather than discrete emotional states (e.g., sadness, fear, anger)”.

EEG-based functional connectivity change was found to be significantly different among emotional states of neutral, positive, or negative. Lee Y-Y, Hsieh S (2014) Classifying Different Emotional States by Means of EEG-Based Functional Connectivity Patterns. PLoS ONE 9(4): e95415. doi.org/10.1371/journal.pone.0095415. A connectivity pattern may be detected by pattern classification analysis using Quadratic Discriminant Analysis. The results indicated that the classification rate was better than chance. They concluded that estimating EEG-based functional connectivity provides a useful tool for studying the relationship between brain activity and emotional states.

Emotions affects learning. Intelligent Tutoring Systems (ITS) learner model initially composed of a cognitive module was extended to include a psychological module and an emotional module. Alicia Heraz et al. introduced an emomental agent. It interacts with an ITS to communicate the emotional state of the learner based upon his mental state. The mental state was obtained from the learner's brainwaves. The agent learns to predict the learners emotions by using machine learning techniques. (Alicia Heraz, Ryad Razaki; Claude Frasson, “Using machine learning to predict learner emotional state from brainwaves” Advanced Learning Technologies, 2007. ICALT 2007. Seventh IEEE International Conference on Advanced Learning Technologies (ICALT 2007)) See also: Ella T. Ma mpusti, Jose S. Ng, Darren James I. Quinto, Grizelda L Teng, Merlin Teodosia C. Suarez, Rhia S. Trogo, “Measuring Academic Affective States of Students via Brainwave Signals”, Knowledge and Systems Engineering (KSE) 2011 Third International Conference on, pp. 226-231, 2011; Judith J. Azcarraga, John Francis Ibanez Jr., Ianne Robert Lim, Nestor Lumanas Jr., “Use of Personality Profile in Predicting Academic Emotion Based on Brainwaves Signals and Mouse Behavior”, Knowledge and Systems Engineering (KSE) 2011 Third International Conference on, pp. 239-244, 2011; Y-Hung Liu, Chien-Te Wu, Yung-Hwa Kao, Ya-Ting Chen, “Single-trial EEG-based emotion recognition using kernel Eigen-emotion pattern and adaptive support vector machine”, Engineering in Medicine and Biology Society (EMBC) 2013 35th Annual International Conference of the IEEE, pp. 4306-4309, 2013, ISSN 1557-170X; Thong Tri Vo, Nam Phuong Nguyen, Toi Vo Van, IFMBE Proceedings, vol. 63, pp. 621, 2018, ISSN 1680-0737, ISBN 978-981-10-4360-4; Adrian Rodriguez Aguinaga, Miguel Angel Lopez Ramirez, Lecture Notes in Computer Science, vol. 9456, pp. 177, 2015, ISSN 0302-9743, ISBN 978-3-319-26507-0; Judith Azcarraga, Merlin Teodosia Suarez, “Recognizing Student Emotions using Brainwaves and Mouse Behavior Data”, International Journa of Distance Education Technologies, vol. 11, pp. 1, 2013, ISSN 1539-3100; OTri Thong Vo, Phuong Nam Nguyen, Van Toi Vo, IFMBE Proceedings, vol. 61, pp. 67, 2017, ISSN 1680-0737, ISBN 978-981-10-4219-5; Alicia Heraz, Claude Frasson, Lecture Notes in Computer Science, vol. 5535, pp. 367, 2009, ISSN 0302-9743, ISBN 978-3-642-02246-3; Hamwira Yaacob, Wahab Abdul, Norhaslinda Kamaruddin, “Classification of EEG signals using MLP based on categorical and dimensional perceptions of emotions”, Information and Communication Technology for the Muslim World (ICT4M) 2013 5th International Conference on, pp. 1-6, 2013; Yuan-Pin Lin, Chi-Hong Wang, Tzyy-Ping Jung, Tien-Lin Wu, Shyh-Kang Jeng, Jeng-Ren Duann, Jyh-Horng Chen, “EEG-Based Emotion Recognition in Music Listening”, Biomedical Engineering IEEE Transactions on, vol. 57, pp. 1798-1806, 2010, ISSN 0018-9294; Yi-Hung Liu, Wei-Teng Cheng, Yu-Tsung Hsiao, Chien-Te Wu, Mu-Derleng, “EEG-based emotion recognition based on kernel Fishers discriminant analysis and spectral powers”, Systems Man and Cybernetics (SMC) 2014 IEEE International Conference on, pp. 2221-2225, 2014.

Using EEG to assess the emotional state has numerous practical applications. One of the first such applications was the development of a travel guide based on emotions by measuring brainwaves by the Singapore tourism group. “By studying the brainwaves of a family on vacation, the researchers drew up the Singapore Emotion Travel Guide, which advises future visitors of the emotions they can expect to experience at different attractions.” (www.lonelyplanet.com/news/2017/04/12/singapore-emotion-travel-guide) Joel Pearson at University of New South Wales and his group developed the protocol of measuring brainwaves of travelers using EEG and decoding specific emotional states.

Another recently released application pertains to virtual reality (VR) technology. On Sep. 18, 2017 Looxid Labs launched a technology that harnesses EEG from a subject waring a VR headset. Looxid Labs intention is to factor in brainwaves into VR applications in order to accurately infer emotions. Other products such as MindMaze and even Samsung have tried creating similar applications through facial muscles recognition. (scoffamyx.com/2017/10/13/looxidlabs-vr-brain-waves-human-emotions/). According to its website (looxidlabs.com/device-2/), the Looxid Labs Development Kit provides a VR headset embedded with miniaturized eye and brain sensors. It uses 6 EEG channels: Fp1, Fp2, AF7, AFB, AR, AF4 in the international 10-20 system.

To assess a users state of mind, a computer may be used to analyze the EEG signals produced by the brain of the user. However, the emotional states of a brain are complex, and the brainwaves associated with specific emotions seem to change overtime. Wei-Long Zheng at Shanghai Liao Tong University used machine learning to identify the emotional brain states and to repeat it reliably. The machine learning algorithm found a set of patterns that clearly distinguished positive, negative, and neutral emotions that worked for different subjects and for the same subjects over time with an accuracy of about 80 percent. (See Wei-Long Zheng, Jia-Yi Zhu, Bao-Liang Lu, Identifying Stable Patterns over Time for Emotion Recognition from EEG, arxiv.org/abs/1601.02197; see also How One Intelligent Machine Learned to Recognize Human Emotions, MIT Technology Review, Jan. 23, 2016.)

MEG: Magnetoencephalography (MEG) is a functional neuroimaging technique for mapping brain activity by recording magnetic fields produced by electrical currents occurring naturally in the brain, using very sensitive magnetometers. Arrays of SQUIDs (superconducting quantum interference devices) are currently the most common magnetometer, while the SERF (spin exchange relaxation-free) magnetometer is being investigated (Hämäläinen, Matti; Han, Riiffa; Ilmoniemi, Risto J.; Knuutila, Jukka; Lounasmaa, Olli V. (1993). “Magnetoencephalography-theory, instrumentation, and applications to noninvasive studies of the working human brain”. Reviews of Modern Physics. 65 (2):413-497. ISSN 0034-6861. doi:10.1103/RevModPhys.65.413.) It is known that “neuronal activity causes local changes in cerebral blood flow, blood volume, and blood oxygenation” (Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. K K Kwong, J. W. Belliveau, D. A. Chester, I. E. Goldberg, R. M. Weisskoff, B. P. Poncelet D. N. Kennedy, B. E. Hoppel, M. S. Cohen, and R. Turner). Using “a 122-channel D.C. SQUID magnetometer with a helmet-shaped detector array covering the subjects head” it has been shown that the “system allows simultaneous recording of magnetic activity all over the head.” (122-channel squid instrument for investigating the magnetic signals from the human brain.) A. I. Ahonen, M. S. Ha malainen, M. J. Kajola, J. E. T. Knuutila, P. P. Laine, O. V. Lounasmaa, L. T. Parkkonen, J. T. Simola, and C. D. Tesche Physica Scripta, Volume 1993, T49A).

In some cases, magnetic fields cancel, and thus the detectable electrical activity may fundamentally differ from the detectable electrical activity obtained via EEG. However, the main types of brain rhythms are detectable by both methods. See: U.S. Pat. Nos. 5,059,814; 5,118,606; 5,136,687; 5,224,203; 5,303,705; 5,325,862; 5,461,699; 5,522,863; 5,640,493; 5,715,821; 5,719,561; 5,722,418; 5,730,146; 5,736,543; 5,737,485; 5,747,492; 5,791,342; 5,816,247; 6,497,658; 6,510,340; 6,654,729; 6,893,407; 6,950,697; 8,135,957; 8,620,206; 8,644,754; 9,118,775; 9,179,875; 9,642,552; 20030018278; 20030171689; 20060293578; 20070156457; 20070259323; 20080015458; 20080154148; 20080229408; 20100010365; 20100076334; 20100090835; 20120046531; 20120052905; 20130041281; 20150081299; 20150262016. See EP1304073A2; EP1304073A3; W02000025668A1; and W02001087153A1.

MEG seek to detect the magnetic dipole emission from an electrical discharge in cells, e.g., neural action potentials. Typical sensors for MEGs are superconducting quantum interference devices (SQUIDs). These currently require cooling to liquid nitrogen or liquid helium temperatures. However, the development of room temperature, or near room temperature superconductors, and miniature cryocoolers, may permit field deployments and portable or mobile detectors. Because MEGs are less influenced by medium conductivity and dielectric properties, and because they inherently detect the magnetic field vector, MEG technology permits volumetric mapping of brain activity and distinction of complementary activity that might suppress detectable EEG signals. MEG technology also supports vector mapping of fields, since magnetic emitters are inherently dipoles, and therefore a larger amount of information is inherently available. See, Reference List Table 3.

EEGs and MEGs can monitor the state of consciousness. For example, states of deep sleep are associated with slower EEG oscillations of larger amplitude. Various signal analysis methods allow for robust identifications of distinct sleep stages, depth of anesthesia, epileptic seizures, and connections to detailed cognitive events.

Neurofeedback: Neurofeedback (NFB), also called neurotherapy or neurobiofeedback, is a type of biofeedback that uses real-time displays of brain activity-most commonly electroencephalography (EEG), to teach self-regulation of brain function. Typically, sensors are placed on the scalp to measure activity, with measurements displayed using video displays or sound. The feedback may be in various other forms as well. Typically, the feedback is sought to be presented through primary sensory inputs, but this is not a limitation on the technique.

The applications of neurofeedback to enhance performance extend to the ads in fields such as music, dance, and acting. A study with conservatoire musicians found that alpha-theta training benefitted the three music domains of musicality, communication, and technique. Historically, alpha-theta training, a form of neurofeedback, was created to assist creativity by inducing hypnagogia, a “borderline waking state associated with creative insights”, through the facilitation of neural connectivity. Alpha-theta training has also been shown to improve novice singing in children. Alpha-theta neurofeedback, in conjunction with heart rate variability training, a form of biofeedback, has also produced benefits in dance by enhancing performance in competitive ballroom dancing and increasing cognitive creativity in contemporary dancers. Additionally, neurofeedback has also been shown to instill a superior flow state in actors, possibly due to greater immersion while performing.

Several studies of brain wave activity in experts while performing a task related to their respective area of expertise revealed certain characteristic telltale signs of so-called “flow” associated with top-flight performance. Mihaly Csikszentmi ha lyi (University of Chicago) found that the most skilled chess players showed less EEG activity in the prefrontal cortex, which is typically associated with higher cognitive processes such as working memory and verbalization, during a game. See, Chris Berka et al., Advanced Brain Monitoring, Carlsbad, EA The International Sport and Society, vol 1, p 87.

Low Energy Neurofeedback System (LENS): The LENS, or Low Energy Neurofeedback System, uses a very low power electromagnetic field, to carry feedback to the person receiving it. The feedback travels down the same wires carrying the brainwaves to the amplifier and computer. Although the feedback signal is weak, it produces a measurable change in the brainwaves without conscious effort from the individual receiving the feedback. The system is software controlled, to receive input from EEG electrodes, to control the stimulation. Through the scalp. Neurofeedback uses a feedback frequency that is different from, but correlates with, the dominant brainwave frequency. When exposed to this feedback frequency, the EEG amplitude distribution changes in power. Most of the time, the brainwaves reduce in power, but at times they also increase in power. In either case, the result is a changed brainwave state and much greater ability for the brain to regulate itself. See, Janice Chen, Content-Based Brainwave Analysis: Memories are not unique. Nature Neuroscience, DOI:10.1038/nn.4450; Andy Coghlan, “Our brains record and remember things in exactly the same way”, New Scientist, Dec. 5, 2016 (www.newscientistcom/adicle/2115093-our-brains-record-and-remember-things-in-exactly-the-same-way/); Brian Pasley, Frontiers in Neuroengineering, doi.org/whb; Helen Thomson, “Hearing our inner voice”. New Scientist, Oct. 29, 2014 (www.newscientistcom/article/mg22429934-000-brain-decoder-can-eavesdrop-on-your-inner-voice/); Bernard Balleine, Proceedings of the National Academy of Sciences, DOI: 10.1073/pnas.1113158108; Wendy Zukerman, “Habits form when brainwaves slow down”, New Scientist, Sep. 26, 2011 (www.newscientistcom/adicle/dn20964-habits-form-when-brainwaves-slow-down/); Smith, K Mind-reading with a brain scan. Nature (2008). oi.org/10.1038/news.2008.650; Kay, K N., Naselaris, T., Prenger, R. J. Er Gallant, J. L Nature advanced online publication doi:10.1038/nature06713 (5 Mar. 2008); Haynes, J.-D. et al. Current Biology 17, 323-328 (2007); Thorn, Catherine A., Hisham Atallah, Mark Howe, and Ann M. Gray biel. “Differential dynamics of activity changes in dorsolateral and dorsomedial striatal loops during learning.” Neuron 66, no. 5 (2010): 781-795; Howe, Mark W., Hisham E. Atallah, Andrew McCool, Daniel J. Gibson, and Ann M. Gray biel. “Habit learning is associated with major shifts in frequencies of oscillatory activity and synchronized spike firing in striatum.” Proceedings of the National Academy of Sciences 108, no. 40 (2011):16801-16806; Pawel Stepien, Wlodzimierz Klonowski and Nikolay Suvorov, Nonlinear analysis of EEG in chess players, EPJ Nonlinear Biomedical Physics 20153:1; Junior, L. R. S., Cesar, F. H. G., Rocha, F. T., and Thomaz, C. E. EEG and Eye Movement Maps of Chess Players. Proceedings of the Sixth International Conference on Pattern Recognition Applications and Methods. (ICPRAM 2017) pp. 343-441. (fei.edu.br/cet/icpram17_Laerciolunior.pdf); You-Yun Lee, Shulan Hsieh. Classifying Different Emotional States by Means of EEG-Based Functional Connectivity Patterns. Apr. 17, 2014, (doi.org/10.1371/journal.pone.0095415). U.S. Pat. No. 9,763,592 provides a system for instructing a user behavior change comprising: collecting and analyzing bioelectrical signal datasets; and providing a behavior change suggestion based upon the analysis. A stimulus may be provided to prompt an action by the user, which may be visual, auditory, or haptic. See also U.S. Pat. No. 9,622,660, 20170041699; 20130317384; 20130317382; 20130314243; 20070173733; and 20070066914. Sensory Stimulation: Light, sound or electromagnetic fields may be used to remotely convey a temporal pattern of brainwaves. See Reference List Table 4.

Light Stimulation: The functional relevance of brain oscillations in the alpha frequency range (8-13 Hz) has been repeatedly investigated through the use of rhythmic visual stimulation. There are two hypotheses on the origin of steady-state visual evoked potential (SSVEP) measured in EEG during rhythmic stimulation: entrainment of brain oscillations and superposition of event-related responses (ERPs). The entrainment but not the superposition hypothesis justifies rhythmic visual stimulation as a means to manipulate brain oscillations because superposition assumes a linear summation of single responses, independent from ongoing brain oscillations. Participants stimulated with the rhythmic flickering light of different frequencies and intensities, and entrainment was measured by comparing the phase coupling of brain oscillations stimulated by rhythmic visual flicker with the oscillations induced by arrhythmic jilt red stimulation, varying the time, stimulation frequency, and intensity conditions. Phase coupling was found to be more pronounced with increasing stimulation intensity as well as at stimulation frequencies closer to each participant's intrinsic frequency. Even in a single sequence of an SSVEP, nonlinear features (intermittency of phase locking) was found that contradict the linear summation of single responses, as assumed by the superposition hypothesis. Thus, evidence suggests that visual rhythmic stimulation entrains brain oscillations, validating the approach of rhythmic stimulation as manipulation of brain oscillations. See, Notbohm A, Kudhs J, Herrmann C S, Modification of Brain Oscillations via Rhythmic Light Stimulation Provides Evidence for Entrainment but Not for Superposition of Event-Related Responses, Front Hum Neurosci. 2016 Feb. 3; 10:10. doi: 10.3389/fnhum.2016.00010.eCollection 2016. It is also known that periodic visual stimulation can trigger epileptic seizures.

It is known to analyze EEG patterns to extract an indication of certain volitional activity (U.S. Pat. No. 6,011,991). This technique describes that an EEG recording can be matched against a stored normalized signal using a computer. This matched signal is then translated into the corresponding reference. The patent application describes a method “a system capable of identifying particular nodes in an individual's brain, the firings of which affect characteristics such as appetite, hunger, thirst, communication skills” and “devices mounted to the person (e.g. underneath the scalp) may be energized in a predetermined manner or sequence to remotely cause particular identified brain node(s) to be fired in order to cause a predetermined feeling or reaction in the individual” without technical description of implementation. This patent also describes, that “brain activity [is monitored] by way of electroencephalograph (EEG) methods, magnetoencephalograph (MEG) methods, and the like.” For example, see U.S. Pat. Nos. 5,816,247 and 5,325,862. See also Reference List Table 5.

Brain Entrainment: Brain entrainment, also referred to as brainwave synchronization and neural entrainment, refers to the capacity of the brain to naturally synchronize its brainwave frequencies with the rhythm of periodic external stimuli, most commonly auditory, visual, or tactile. Brainwave entrainment technologies are used to induce various brain states, such as relaxation or sleep, by creating stimuli that occur at regular, periodic intervals to mimic electrical cycles of the brain during the desired states, thereby “training” the brain to consciously alter states. Recurrent acoustic frequencies, flickering lights, or tactile vibrations are the most common examples of stimuli applied to generate different sensory responses. It is hypothesized that listening to these beats of certain frequencies one can induce a desired state of consciousness that corresponds with specific neural activity. Patterns of neural firing, measured in Hz, correspond with alertness states such as focused attention, deep sleep, etc.

Neural oscillations are rhythmic or repetitive electrochemical activity in the brain and central nervous system. Such oscillations can be characterized by their frequency, amplitude, and phase. Neural tissue can generate oscillatory activity driven by mechanisms within individual neurons, as well as by interactions between them. They may also adjust frequency to synchronize with the periodic vibration of external acoustic or visual stimuli. The functional role of neural oscillations is still not fully understood; however, they have been shown to correlate with emotional responses, motor control, and a number of cognitive functions including information transfer, perception, and memory. Specifically, neural oscillations, in particular theta activity, are extensively linked to memory function, and coupling between theta and gamma activity is considered to be vital for memory functions, including episodic memory. Electroencephalography (EEG) has been most widely used in the study of neural activity generated by large groups of neurons, known as neural ensembles, including investigations of the changes that occur in electroencephalographic profiles during cycles of sleep and wakefulness. EEG signals change dramatically during sleep and show a transition from faster frequencies to increasingly slower frequencies, indicating a relationship between the frequency of neural oscillations and cognitive states, including awareness and consciousness.

The term “entrainment” has been used to describe a shared tendency of many physical and biological systems to synchronize their periodicity and rhythm through interaction. This tendency has been identified as specifically pertinent to the study of sound and music generally, and acoustic rhythms specifically. The most ubiquitous and familiar examples of neuromotor entrainment to acoustic stimuli are observable in spontaneous foot or finger tapping to the rhythmic beat of a song. Exogenous rhythmic entrainment, which occurs outside the body, has been identified and documented for a variety of human activities, which include the way people adjust the rhythm of their speech patterns to those of the subject with whom they communicate, and the rhythmic unison of an audience dapping. Even among groups of strangers, the rate of breathing, locomotive, and subtle expressive motor movements, and rhythmic speech patterns have been observed to synchronize and entrain, in response to an auditory stimulus, such as a piece of music with a consistent rhythm. Furthermore, motor synchronization to repetitive tactile stimuli occurs in animals, including cats and monkeys as well as humans, with accompanying shifts in electroencephalogram (EEG) readings. Examples of endogenous entrainment, which occurs within the body, include the synchronizing of human circadian sleep-wake cycles to the 24-hour cycle of light and dark, and the frequency following response of humans to sounds and music.

Brainwaves, or neural oscillations, share the fundamental constituents with acoustic and optical waves, including frequency, amplitude, and periodicity. The synchronous electrical activity of cortical neural ensembles can synchronize in response to external acoustic or optical stimuli and also entrain or synchronize their frequency and phase to that of a specific stimulus. Brainwave entrainment is a colloquialism for such ‘neural entrainment’, which is a term used to denote the way in which the aggregate frequency of oscillations produced by the synchronous electrical activity in ensembles of cortical neurons can adjust to synchronize with the periodic vibration of an external stimuli, such as a sustained acoustic frequency perceived as pitch, a regularly repeating pattern of intermittent sounds, perceived as rhythm, or of a regularly rhythmically intermittent flashing light.

Changes in neural oscillations, demonstrable through electroencephalogram (EEG) measurements, are precipitated by listening to music, which can modulate autonomic arousal ergotropically and trophotropically, increasing and decreasing arousal respectively. Musical auditory stimulation has also been demonstrated to improve immune function, facilitate relaxation, improve mood, and contribute to the alleviation of stress.

The Frequency following response (FFR), also referred to as Frequency Following Potential (FFP), is a specific response to hearing sound and music, by which neural oscillations adjust their frequency to match the rhythm of auditory stimuli. The use of sound with intent to influence cortical brainwave frequency is called auditory driving, by which frequency of neural oscillation is ‘driven’ to entrain with that of the rhythm of a sound source. See Reference List Table 6.

Baseline correction of event-related time-frequency measure may be made by taking pre-event baseline activity into consideration. A baseline period is defined by averaging values within a time window. Methods for baseline correction in time-frequency analysis include various baseline value normalizations. The question of whether different emotional states are associated with specific patterns of physiological response has long being a subject of neuroscience research. See, Reference List Table 7.

Electroencephalograms (EEG) and functional Magnetic Resonance Imaging, fMRI have been used to study specific brain activity associated with different emotional states. Mauss and Robinson, in their review paper, have indicated that “emotional state is likely to involve circuits rather than any brain region considered in isolation” (Mauss I B, Robinson M D (2009) Measures of emotion: A review. Cogn Emot 23: 209-237.) The amplitude, latency from the stimulus, and covariance (in the case of multiple electrode sites) of each component can be examined in connection with a cognitive task (ERP) or with no task (EP). Steady-state visually evoked potentials (551/EPs) use a continuous sinusoidally-modulated flickering light, typically superimposed in front of a TV monitor displaying a cognitive task. The brain response in a narrow frequency band containing the stimulus frequency is measured. Magnitude, phase, and coherence (in the case of multiple electrode sites) may be related to different parts of the cognitive task. Brain entrainment may be detected through EEG or MEG activity. Brain entrainment may be detected through EEG or MEG activity. See Reference List Table 8.

The entrainment hypothesis (Thut and Miniussi, 2009; Thut et al., 2011a, 2012), suggests the possibility of inducing a particular oscillation frequency in the brain using an external oscillatory force (e.g., rTMS, but also tACS). The physiological basis of oscillatory cortical activity lies in the timing of the interacting neurons; when groups of neurons synchronize their firing activities, brain rhythms emerge, network oscillations are generated, and the basis for interactions between brain areas may develop (Buzsaki, 2006). Because of the variety of experimental protocols for brain stimulation, limits on descriptions of the actual protocols employed, and limited controls, consistency of reported studies is lacking, and extrapolability is limited. Thus, while there is various consensus in various aspects of the effects of extracranial brain stimulation, the results achieved have a degree of uncertainty dependent on details of implementation. On the other hand, within a specific experimental protocol, it is possible to obtain statistically significant and repeatable results. This implies that feedback control might be effective to control implementation of the stimulation for a given purpose; however, studies that employ feedback control are lacking.

Different cognitive states are associated with different oscillatory patterns in the brain (Buzsaki, 2006; Canopy and Knight 2010; Varela et al., 2001). Thut et al. (2011b) directly tested the entrainment hypothesis by means of a concurrent EEG-TMS experiment. They first determined the individual source of the parietal-occipital alpha modulation and the individual alpha frequency (magnetoencephalography study). They then applied rTMS at the individual alpha power while recording the EEG activity at rest. The results confirmed the three predictions of the entrainment hypothesis: the induction of a specific frequency after TMS, the enhancement of oscillation during TMS stimulation due to synchronization, and phase alignment of the induced frequency and the ongoing activity (Thut et al., 2011b). If associative stimulation is a general principle for human neural plasticity in which the timing and strength of activation are critical factors, it is possible that synchronization within or between areas using an external force to phase/align oscillations can also favor efficient communication and associative plasticity (or alter communication). In this respect associative, cortico-cortical stimulation has been shown to enhance the coherence of oscillatory activity between the stimulated areas (Plewnia et al., 2008). In a coherence resonance (Longtin, 1997), the addition of a certain amount of noise in an excitable system results in the most coherent and proficient oscillatory responses. The brain's response to external timing-embedded stimulation can result in a decrease in phase variance and an enhanced alignment (clustering) of the phase components of the ongoing EEG activity (entraining, phase resetting) that can change the signal-to-noise ratio and change signal efficacy.

If one considers neuron activity within the brain as a set of loosely coupled oscillators, then the various parameters that might be controlled include the size of the region of neurons, frequency of oscillation, resonant frequency or time-constant, oscillator damping, noise, amplitude, coupling to other oscillators, and of course, external influences that may include stimulation and/or power loss. In a human brain, pharmacological intervention may be significant. For example, drugs that alter excitability, such as caffeine, neurotransmitter release and reuptake, nerve conductance, etc. can all influence operation of the neural oscillators. Likewise, sub-threshold external stimulation effects, including DC, AC, and electromagnetic effects, can also influence the operation of the neural oscillators.

Phase resetting or shifting can synchronize inputs and favor communication and, eventually, Hebbian plasticity (Hebb, 1949). Thus, rhythmic stimulation may induce a statistically higher degree of coherence in spiking neurons, which facilitates the induction of a specific cognitive process (or hinders that process). Here, the perspective is slightly different (coherence resonance), but the underlining mechanisms are similar to the ones described so far (stochastic resonance), and the additional key factor is the repetition at a specific rhythm of the stimulation. See, Cade “The Awakened Mind: Biofeedback and the Development of Higher States of Awareness” (Dell, 1979); Anna Wise, “The High-Performance Mind: Mastering Brainwaves for Insight, Healing, and Creativity”.

Entrainment is plausible because of the characteristics of the demonstrated EEG responses to a single TMS pulse, which have a spectral composition which resembles the spontaneous oscillations of the stimulated cortex. For example, TMS of the “resting” visual (Rosa nova et al., 2009) or motor cortices (Veniero et al., 2011) triggers alpha-waves, the natural frequency at the resting state of both types of cortices. With the entrainment hypothesis, the noise generation framework moves to a more complex and extended level in which noise is synchronized with on-going activity. Nevertheless, the model to explain the outcome will not change, stimulation will interact with the system, and the final result will depend on introducing or modifying the noise level. The entrainment hypothesis makes clear predictions with respect to online repetitive TMS paradigms' frequency engagement as well as the possibility of inducing phase alignment, i.e., a reset of ongoing brain oscillations via external spTIVS (Thut et al., 2011a, 2012; Veniero et al., 2011). The entrainment hypothesis is superior to the localization approach in gaining knowledge about how the brain works, rather than where or when a single process occurs. TMS pulses may phase-align the natural, ongoing oscillation of the target cortex. When additional TMS pulses are delivered in synchrony with the phase-aligned oscillation (i.e., at the same frequency), further synchronized phase-alignment will occur, which will bring the oscillation of the target area in resonance with the TMS train. Thus, entrainment may be expected when TMS is frequency-tuned to the underlying brain oscillations (Veniero et al., 2011).

Binaural Beats: Binaural beats are auditory brainstem responses which originate in the superior olivary nucleus of each hemisphere. They result from the interaction of two different auditory impulses, originating in opposite ears, below 1000 Hz and which differ in frequency between one and 30 Hz. For example, if a pure tone of 400 Hz is presented to the right ear and a pure tone of 410 Hz is presented simultaneously to the left ear, an amplitude modulated standing wave of 10 Hz, the difference between the two tones, is experienced as the two wave forms mesh in and out of phase within the superior olivary nuclei. This binaural beat is not heard in the ordinary sense of the word (the human range of hearing is from 20-20,000 Hz). It is perceived as an auditory beat and theoretically can be used to entrain specific neural rhythms through the frequency-following response (FFR)—the tendency for cortical potentials to entrain to or resonate at the frequency of an external stimulus. Thus, it is theoretically possible to utilize a specific binaural-beat frequency as a consciousness management technique to entrain a specific cortical rhythm. The binaural-beat appears to be associated with an electroencephalographic (EEG) frequency-following response in the brain.

Uses of audio with embedded binaural beats that are mixed with music or various pink or background sound are diverse. They range from relaxation, meditation, stress reduction, pain management, improved sleep quality, decrease in sleep requirements, super learning, enhanced creativity and intuition, remote viewing, telepathy, and out-of-body experience and lucid dreaming. Audio embedded with binaural beats is often combined with various meditation techniques, as well as positive affirmations and visualization.

When signals of two different frequencies are presented, one to each ear, the brain detects phase differences between these signals. “Under natural circumstances, a detected phase difference would provide directional information. The brain processes this anomalous information differently when these phase differences are heard with stereo headphones or speakers. A perceptual integration of the two signals takes place, producing the sensation of a third “beat” frequency. The difference between the signals waxes and wanes as the two different input frequencies mesh in and out of phase. As a result of these constantly increasing and decreasing differences, an amplitude-modulated standing wave—the binaural beat—is heard. The binaural beat is perceived as a fluctuating rhythm at the frequency of the difference between the two auditory inputs. Evidence suggests that the binaural beats are generated in the brainstem's superior olivary nucleus, the first site of contralateral integration in the auditory system. Studies also suggest that the frequency-following response originates from the inferior colliculus. This activity is conducted to the cortex where it can be recorded by scalp electrodes. Binaural beats can easily be heard at the low frequencies (<30 Hz) that are characteristic of the EEG spectrum.

Synchronized brainwaves have long been associated with meditative and hypnogogic states, and audio with embedded binaural beats has the ability to induce and improve such states of consciousness. The reason for this is physiological. Each ear is “hardwired” (so to speak) to both hemispheres of the brain. Each hemisphere has its own olivary nucleus (sound-processing center) which receives signals from each ear. In keeping with this physiological structure, when a binaural beat is perceived there are actually two standing waves of equal amplitude and frequency present, one in each hemisphere. So, there are two separate standing waves entraining portions of each hemisphere to the same frequency. The binaural beats appear to contribute to the hemispheric synchronization evidenced in meditative and hypnogogic states of consciousness. Brain function is also enhanced through the increase of cross-colossal communication between the left and right hemispheres of the brain. See Reference List Table 9.

Isochronic Tones: Isochronic tones are regular beats of a single tone that are used alongside monaural beats and binaural beats in the process called brainwave entrainment. At its simplest level, an isochronic tone is a tone that is being turned on and off rapidly. They create sharp, distinctive pulses of sound. See Reference List Table 10.

Time-Frequency Analysis: Brian J. Roach and Daniel H. Mathalon, “Event-related EEG time-frequency analysis: an overview of measures and analysis of early gamma band phase locking in schizophrenia. Schizophrenia Bull. USA. 2008; 34:5:907-926., describes a mechanism for EEG time-frequency analysis. Fourier and wavelet transforms (and their inverse) may be performed on EEG signals. See Reference List Table 11.

There are many approaches to time-frequency decomposition of EEG data, including the short-term Fourier transform (STFT), (Gabor D. Theory of Communication. J. Inst. Electr. Engrs.1946; 93:429-457) continuous (Daubechies I. Ten Lectures on Wavelets. Philadelphia, Pa.: Society for Industrial and Applied Mathematics; 1992:357.21. Combes J M, Grossmann A, Tchamitchian P. Wavelets: Time-Frequency Methods and Phase Space-Proceedings of the International Conference; Dec. 14-18, 1987; Marseille, France) or discrete (Ma Ilat 5G. A theory for multi resolution signal decomposition: the wavelet representation. IEEE Trans Pattern Anal Mach Inte11.1989; 11:674-693) wavelet transforms, Hilbert transform (Lyons R G. Understanding Digital Signal Processing. 2nd ed. Upper Saddle River, N.J.: Prentice Hall PTR; 2004:688), and matching pursuits (Mallat S, Zhang Z. Matching pursuits with time-frequency dictionaries. IEEE Trans. Signal Proc 1993; 41(12)3397-3415). Prototype analysis systems may be implemented using, for example, MatLab with the Wavelet Toolbox, www.mathworks.com/products/wavelet html. See Reference List Table 12.

Single instruction, multiple data processors, such as graphics processing units including the nVidia CUDA environment or AMD Firepro high-performance computing environment are known, and may be employed for general purpose computing, finding particular application in data matrix transformations. See Reference List Table 13.

Statistical analysis may be presented in a form that permits parallelization, which can be efficiently implemented using various parallel processors, a common form of which is a SIMD (single instruction, multiple data) processor, found in typical graphics processors (GPUs). Artificial neural networks have been employed to analyze EEG signals. See Reference List Table 14.

Principal Component Analysis: Principal component analysis (PCA) is a statistical procedure that uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of linearly uncorrelated variables called principal components. If there are n observations with p variables, then the number of distinct principal components is min(n−1,p). This transformation is defined in such a way that the first principal component has the largest possible variance (that is, accounts for as much of the variability in the data as possible), and each succeeding component in turn has the highest variance possible under the constraint that it is orthogonal to the preceding components. The resulting vectors are an uncorrelated orthogonal basis set. This is useful for segregating components of a signal into self-correlated groups, and to segregate uncorrelated groups. PCA is sensitive to the relative scaling of the original variables. PCA is the simplest of the true eigenvector-based multivariate analyses. Often, its operation can be thought of as revealing the internal structure of the data in a way that best explains the variance in the data. If a multivariate dataset is visualized as a set of coordinates in high-dimensional data space (1 axis per variable), PCA can supply the user with a lower-dimensional picture, a projection of this object when viewed from its most informative viewpoint. This is done by using only the first few principal components so that the dimensionality of the transformed data is reduced. PCA is closely related to factor analysis. Factor analysis typically incorporates more domain specific assumptions about the underlying structure and solves eigenvectors of a slightly different matrix. PCA is also related to canonical correlation analysis (CCA). CCA defines coordinate systems that optimally describe the cross-covariance between two datasets while PCA defines a new orthogonal coordinate system that optimally describes variance in a single dataset. See, en.wikipedia.org/wiki/Principal component analysis.

A general model for confirmatory factor analysis is expressed as x=α+Λξ+ε. The covariance matrix is expressed as E[(x−μ)(x−μ)′]=ΛΦΛ′+Θ. If residual covariance matrix Θ=0 and correlation matrix among latent factors Φ=I, then factor analysis is equivalent to PCA and the resulting covariance matrix is simplified to Σ=ΛΛ′. When there are p number of variables and all p components (or factors) are extracted, this covariance matrix can alternatively be expressed into Σ=DΛD′, or Σ=λDAD′, where D=n×p orthogonal matrix of eigenvectors, and Λ=λA, p×p matrix of eigenvalues, where λ is a scalar, and A is a diagonal matrix whose elements are proportional to the eigenvalues of Σ. The following three components determine the geometric features of the observed data: parameterizes the volume, D indicates the orientation, and A represents the shape of the observation.

When population heterogeneity is explicitly hypothesized as in model-based cluster analysis, the observed covariance matrix is decomposed into the following general form Σ_k=λ_kD_kA_kD_k^T, where λ_k parameterizes the volume of the k^th duster, D_k indicates the orientation of that duster, and A_k represents the shape of that duster. The subscript k indicates that each component (or duster) can have different volume, shape, and orientation. Assume a random vector X, taking values ^m, has a mean and covariance matrix of μ_x and Σ_X, respectively. λ₁>λ₂> . . . >λ_m>0 are ordered eigenvalues of Σ_X, such that the i^th eigenvalue of Σ_X means the i^th largest of them. Similarly, a vector α_i is the i^th eigenvector of Σ_X when it corresponds to the eigenvalue of Σ_X. To derive the form of principal components (PCs), consider the optimization problem of maximizing var[α₁^T X]=α₁^TΣ_Xα₁, subject to α₁^Tα₁=1. The Lagrange multiplier method is used to solve this question.

$L\left({\alpha }_1,{\phi }_1\right)={\alpha }_1^T{\sum }_X{\alpha }_1+{\phi }_1\left({\alpha }_1^T{\alpha }_1-1\right),\frac{\partial L}{\partial {\alpha }_1}=2{\sum }_X{\alpha }_12{\phi }_1{\alpha }_1=0\Rightarrow {\sum }_X{\alpha }_1=-{\phi }_1{\alpha }_1\Rightarrow \mathrm{var}\left[{\alpha }_1^TX\right]=-{\phi }_1{\alpha }_1^T{\alpha }_1=-{\phi }_{1.}$

Because −ϕ₁ is the eigenvalue of Σ_X, with α₁ being the corresponding normalized eigenvector, var[α₁^T X] is maximized by choosing α₁ to be the first eigenvector of Σ_X. In this case, ₁=α₁^T X is named the first PC of X, α₁ is the vector of coefficients for 1, and var(₁)=λ₁.

To find the second PC, ₂=α₂^T X, we need to maximize var[α₂^T X]=α₂^T Σ_X α₂ subject to ₂ being uncorrelated with ₁. Because cov(α₁^T X, α₂¹ X)=0⇒α₁^TΣ_X α₂=0⇒α₁^T α₂=0, this problem is equivalently set as maximizing α₂^T Σ_Xα₂, subject to a₁^T α₂=0, and α₂^T α₂=1. We still make use of the Lagrange multiplier method

$L\left({\alpha }_2,{\phi }_1,{\phi }_2\right)={\alpha }_2^T\sum _X{\alpha }_2+{\phi }_1{\alpha }_1^T{\alpha }_2+{\phi }_2\left({\alpha }_2^T{\alpha }_2-1\right)\frac{\sigma L}{\partial {\alpha }_2}=2\sum _X{\alpha }_2+{\phi }_1{\alpha }_1+2{\phi }_2{\alpha }_2=0\Rightarrow {\alpha }_1^T\left(2\sum _X{\alpha }_2+{\phi }_1{\alpha }_1+2{\phi }_2{\alpha }_2\right)=0\Rightarrow {\phi }_1=0\Rightarrow \sum _X{\alpha }_2=-{\phi }_2{\alpha }_2\Rightarrow {\alpha }_2^T\sum _X{\alpha }_2=-{\phi }_2.$

Because −ϕ₂ is the eigenvalue of Σ_X, with α₂ being the corresponding normalized eigenvector, var[α₂^T X] is maximized by choosing α₂ to be the second eigenvector of Σ_X. In this case, ₂=α₂^T X is named the second PC of X, α₂ is the vector of coefficients for ₂, and var(₂)=λ₂. Continuing in this way, it can be shown that the i-th PC _i^=α_i^T X is constructed by selecting α_i to be the i^th eigenvector of Σ_X, and has a variance of λ_i. The key result in regards to PCA is that the principal components are the only set of linear functions of original data that are uncorrelated and have orthogonal vectors of coefficients.

For any positive integer p≤m, let B=[β₁, β₂, . . . , β_p] be an real m×p matrix with orthonormal columns, i.e., β_i^T β_j=δ_ij, and Y=B^T X. Then the trace of covariance matrix of Y is maximized by taking B=[α₁, α₂, . . . , α_p], where α_i is the i-th eigenvector of Σ_X. Because Σ_X is symmetric with all distinct eigenvalues so {α₁, β₂, . . . , φ_m} is an orthonormal basis with α_i being the i-th eigenvector of Σ_X, and we can represent the columns of B as

${\beta }_i=\sum _{j=1}c_{\mathrm{ji}}{\alpha }_j,i=1,\dots ,p,$

So we have B=PC, where P=[α₁, . . . , α_m], C={c_ij} is an m×p matrix. Then, P^T ρ_X P=Λ, with Λ being a diagonal matrix whose k-th diagonal element is λ_k, and the covariance matrix of Y is,

Σ_Y=B^T Σ_X B=C^T P^T Σ_X PC=C^T ΛC=λ₁ c₁ c₁^T+ . . . + λ_m c_m c_m^T, where c_i^T is the i-th row of C. So,

$\mathrm{trace}\left(\sum _Y\right)=\sum _{i=1}^m{\lambda }_i\mathrm{trace}\left(c_ic_i^T\right)=\sum _{i=1}^m{\lambda }_i\mathrm{trace}\left(c_i^Tc_i\right)=\sum _{i=1}^m{\lambda }_ic_i^Tc_i=\sum _{i=1}^m\left(\sum _{j=1}^pc_{\mathrm{ij}}^2\right){\lambda }_i.$

Because C^T C=B^T PP^T B=B^T B=I, so

$\mathrm{trace}\left(C^TC\right)=\sum _{i=1}^m\sum _{j=1}^Vc_{\mathrm{ij}}^2=p,$

and the columns of C are orthonormal.
By the Gram-Schmidt method, C can expand to D, such that D has its columns as an orthonormal basis of ^m and contains C as its first p columns. D is square shape, thus being an orthogonal matrix and having its rows as another orthonormal basis of ^m. One row of C is a part of one row of D, so

$\sum _{j=1}^pc_{\mathrm{ij}}^2\le 1,i=1,\dots ,m.$

Considering the constraints

$\sum _{j=1}^pc_{\mathrm{ij}}^2\le 1,\sum _{i=1}^m\sum _{j=1}^pc_{\mathrm{ij}}^2=p$

and the objective

$\sum _{i=1}^m\left(\sum _{j=1}^pc_{\mathrm{ij}}^2\right){\lambda }_i.$

We derive that trace(Σ_Y) is maximized if

$\sum _{j=1}^pc_{\mathrm{ij}}^2=1$

for i=1, . . . , p, and

$\sum _{j=1}^pc_{\mathrm{ij}}^2=0$

for i=p+1 . . . , m. When B=[α₁, β₂, . . . , α_p], straightforward calculation yields that C is an all zero matrix except c_ii=1, i=1, . . . , p. This fulfills the maximization condition. Actually, by taking B=[γ₁, γ₂, γ_p], where {γ₁, γ₂, γ_p} is any orthonormal basis of the subspace of span{α₁, α₂, . . . , α_p}, the maximization condition is also satisfied, yielding the same trace of the covariance matrix of Y.

Suppose that we wish to approximate the random vector X by its projection onto a subspace spanned by columns of B, where B=[β₁, γ₂, . . . , β_m], is a real m×p matrix with orthonormal columns, i.e., β_i^T β_j=δ_ij. If σ_i² is the residual variance for each component of X, then

$\sum _{i=1}^m{\sigma }_i^2$

is minimized if B=[α₁, α₂, . . . , α_p], where {α₁, α₂, . . . , β_p} are the first p eigenvectors of Σ_X. In other words, the trace of the covariance matrix of X−BB^T X is minimized if B=[α₁, α₂, . . . , α_p]. When E(X)=0, which is a commonly applied preprocessing step in data analysis methods, this property is saying that E∥X−BB^T X∥² is minimized if B=[α₁, α₂, . . . , α_p]. The projection of a random vector X onto a subspace spanned by columns of B is {circumflex over (X)}=BB^T X. Then the residual vector is ε=X−BB^T X, which has a covariance matrix Σ_ε=(I−BB^T)Σ_X(I−BB^T). Then,

$\sum _{i=1}^m{\sigma }_i^2=\mathrm{trace}\left(\sum _{\varepsilon }\right)=\mathrm{trace}\left(\sum _X-\sum _X{\mathrm{BB}}^T-{\mathrm{BB}}^T\sum _X+{\mathrm{BB}}^T\sum _X{\mathrm{BB}}^T\right).$

Also, we know:

trace(Σ_X BB^T)=trace(BB^T Σ_X)=trace(B^T Σ_X B)

trace(BB^T Σ_x BB^T)=trace(B^T Σ_X BB^T B)=trace(B^T Σ_X B)

The last equation comes from the fact that B has orthonormal columns. So,

$\sum _{i=1}^m{\sigma }_i^2=\mathrm{trace}\left({\sum }_X\right)-\mathrm{trace}\left(B^T{\sum }_XB\right).$

To minimize

$\sum _{i=1}^m{\sigma }_i^2,$

it suffices to maximize trace(B^T Σ_X B). This can be done by choosing B=[α₁, α₂, . . . , α_p], where {α₁, α₂, . . . , α_p} are the first p eigenvectors of Σ_X, as above.

See, Pietro Amenta, Luigi D'Ambra, “Generalized Constrained Principal Component Analysis with External Information,” (2000). We assume that data on K sets of explanatory variables and S criterion variables of n statistical units are collected in matrices X_k (k=1, . . . K) and Y_s (s=1, . . . , S) of orders (n×p₁), . . . , (n×p_K) and (n×q₁), . . . , (n×q_s), respectively. We suppose, without loss of generality, identity matrices for the metrics of the spaces of variables of X_k and Y_s with D_n=diag (1/n), weight matrix of statistical units. We assume, moreover, that X_k's and Y_s's are centered as to the weights D_n.

Let X=[X₁| . . . |X_K] and Y=[Y₁| . . . |Y_s], respectively, be K and S matrices column linked to orders (n=Σ_kp_k) and (n×Σ_sq_s). Let be, also, W_Y=YY′ while we denote ν_k the coefficients vector (p_k,1) of the linear combination for each X_k such that k=X_kν_k. Let C_k be the matrix of dimension p_k×m (m≤p_k), associated to the external information explanatory variables of set k.

Generalized CPCA (GCPCA) (Amenta, D'Ambra, 1999) with external information consists in seeking for K coefficients vectors ν_k (or, in same way, K linear combinations _k) subject to the restriction C_k′ν_k=0 simultaneously, such that:

$\begin{array}{cc}\{\begin{array}{c}\max \sum _{i=1}^K\sum _{j=1}^K〈Y^{\prime }X_iv_i,Y^{\prime }X_jv_j〉\\ \mathrm{with}\mathrm{the}\mathrm{constraints}\begin{array}{c}\sum _{k=1}^K{X_kv_k}^2=1\\ \sum _{k=1}^KC_k^{\prime }v_k=0\end{array}\end{array}& \left(1\right)\end{array}$

or, in equivalent way,

$\{\begin{array}{c}\max v^{\prime }\left(A^{\prime }A\right)v\\ \mathrm{with}\mathrm{the}\mathrm{constraints}\begin{array}{c}{v}^{\prime }\mathrm{Bv}=1\\ C^{\prime }v=0\end{array}\end{array}\mathrm{or}\{\begin{array}{c}\max f^{\prime }B^{-0.5}A^{\prime }{\mathrm{AB}}^{-0.5}f\\ \mathrm{with}\mathrm{the}\mathrm{constraints}\begin{array}{c}{f}^{\prime }f=1\\ C^{\prime }v=0\end{array}\end{array}$

where A=Y′X, B=diag(X₁′X, . . . , X_K′X_K), C′=[C₁′| . . . |C₅′], ν′=(ν₁′| . . . |ν_k′) and f=B^0.5νwith

$A^{\prime }A=\left[\begin{array}{ccc}{X}_1^{\prime }{\mathrm{YY}}^{\prime }X_1& \dots & X_1^{\prime }{\mathrm{YY}}^{\prime }X_K\\ ⋮& \ddots & ⋮\\ X_K^{\prime }{\mathrm{YY}}^{\prime }X_1& \dots & X_k^{\prime }{\mathrm{YY}}^{\prime }X_k\end{array}\right].$

The constrained maximum problem turns out to be an extension of criterion sup_{Σk∥k∥2=1}Σ_iΣ_ki, _k (Sabatier, 1993) with more sets of criterion variables with external information. The solution of this constrained maximum problem leads to solve the eigen-equation

(P_X−P_XB−1C)W_Yg=λg,

where g=Xν, P_X−P_XB−1C=Σ_k=1(P_Xk−P_{Xk(Xk′Xk)−1Ck}) is the oblique projector operator associated to the direct sum decomposition of _n, ⁿ=Im(P_X−P_XB−1C){dot over (⊕)}Im(P_C){dot over (⊕)}Ker(P_X) with P_Xk=X_k(X_k′X_k)⁻¹ X_k′ and P_C=C(C′B⁻¹C)⁻¹C′B⁻¹, respectively, / and B⁻¹ orthogonal projector operators onto the subspaces spanned by the columns of matrices X_k and C. Furthermore, P_XB−1C=XB⁻¹C(C′ B⁻¹ C)⁻¹C′ B⁻¹ X′ is the orthogonal projector operator onto the subspace spanned the columns of the matrix XB⁻¹C. Starting from the relation

(P_Xk−P_{Xk(Xk′Xk)−1Ck})W_Yg=λX_kν_k

(which is obtained from the expression (I−P_C)X′W_Yg=λBν) the coefficients vectors ν_k and the linear combinations _k=X_kν_k maximizing (1) can be given by the relations

$v_k=\frac{1}{\lambda }{\left(X_k^{\prime }X_k\right)}^{-1}\left(I-P_{C_k}\right)X_k^{\prime }W_Y\mathrm{Xv}\mathrm{and}$ $z_k=\frac{1}{\lambda }\left(P_{X_k}-P_{{X_k\left(X_k^{\prime }X_k\right)}^{-1}C_k}\right)W_Y\mathrm{Xv},$

respectively.

The solution eigenvector g can be written, as the sum of the linear combinations _k:g=Σ_kX_kν_k. Notice that the eigenvalues associated to the eigen-system are, according to the Sturm theorem, lower or equal than those of GCPCA eigen system: Σ_k=1^KP_Xk W_Y g=λg, See Reference List Table 15.

Spatial Principal Component Analysis

Let J(t, i;α, s) be the current density in voxel i, as estimated by LORETA, in condition a at t time-frames after stimulus onset for subject s. Let area:Voxel→fBA be a function, which assigns to each voxel i∈Voxel the corresponding fBAb∈fBA. In a first pre-processing step, we calculate for each subjects the value of the current density averaged over each Fba

$\begin{array}{cc}x\left(t,b;\alpha ,s\right)=\frac{1}{N_b}\sum _{i\in b}J\left(t,i;\alpha ,s\right)& \left(4\right)\end{array}$

where N_b is the number of voxels in the fBAb, in condition α for subjects.

In the second analysis stage, the mean current density x(t, b;α, s) from each fBAb, for every subjects and conditionoα, was subjected to spatial PCA analysis of the correlation matrix and varimax rotation

In the present study, the spatial PCA uses the above-defined fBAs as variables sampled along the time epoch for which EEG has been sampled (0-1000 ms; 512 time-frames), and the inverse solution was estimated. Spatial matrices (each matrix was sized b×t=36×512 elements) for every subject and condition were collected, and subjected to PCA analyses, including the calculation of the covariance matrix; eigenvalue decomposition and varimax rotation, in order to maximize factor loadings. In other words, in the spatial PCA analysis, we approximate the mean current density for each subject in each condition as (2)

$x\left(t;\alpha ,s\right)\approx x_0\left(\alpha ,s\right)+\sum _kc_k\left(t\right)x_k\left(\alpha ,s\right),$

where here x(t;α, s)∈R³⁶ is a vector, which denotes the time-dependent activation of the fBAs, x₀(α, s) is their mean activation, and x_k(α, s) and c_k are the principal components and their corresponding coefficients (factor loadings) as computed using the principal PCA.

See, download.|ww.com/wolterskluwer.com/WNR_1_1_2010_03_22_ARZY_1_SDCl.doc.

Singular spectrum analysis (SSA): SSA is a nonparametric spectral estimation method. It combines elements of classical time series analysis, multivariate statistics, multivariate geometry, dynamical systems, and signal processing. SSA can be an aid in the decomposition of time series into a sum of components, each having a meaningful interpretation. The name “singular spectrum analysis” relates to the spectrum of eigenvalues in a singular value decomposition of a covariance matrix, and not directly to a frequency domain decomposition. (see en.wikipeda.org/wiki/Singular_spectrum_analysis.)

In practice, SSA is a nonparametric spectral estimation method based on embedding a time series {X(t): t=1, . . . N} in a vector space of dimension M. SSA proceeds by diagonalizing the M×M lag-covariance matrix C_X of X(t) to obtain spectral information on the time series, assumed to be stationary in the weak sense. The matrix C_X can be estimated directly from the data as a Toeplitz matrix with constant diagonals, i.e., its entries C_i depend only on the lag |i−j|:

$c_{ij}=\frac{1}{N-i-j}\sum _{t=1}^{N-\left|i-j\right|}X\left(t\right)X\left(t+i-j\right)$

An alternative way to compute C_X, is by using the N′×M “trajectory matrix” D that is formed by M lag-shifted copies of X(t), which are N′=N−M+1 long; then

$C_X=\frac{1}{N^{\prime }}D^tD$

The M eigenvectors E_k of the lag-covariance matrix C_X are called temporal empirical orthogonal functions (EOFs). The eigenvalues λ_k of C_X account for the partial variance in the direction E_k and the sum of the eigenvalues, i.e., the trace of C_A gives the total variance of the original time series X(t). The name of the method derives from the singular values λ_k^1/2 of C_X.

Projecting the time series onto each EOF yields the corresponding temporal principal components (PCs) IV

$A_k\left(t\right)=\sum _{j=1}^MX\left(t+j-1\right)E_k\left(j\right).$

An oscillatory mode is characterized by a pair of nearly equal SSA eigenvalues and associated PCs that are in approximate phase quadrature, which can efficiently represent a nonlinear, nonharmonic oscillation, due to the fact that a single pair of data-adaptive SSA eigenmodes often will capture the basic periodicity of an oscillatory mode than fixed basis functions, such as the sines and cosines used in the Fourier transform.

The window width M determines the longest periodicity captured by SSA. Signal-to-noise separation can be obtained by merely inspecting the slope break in a “scree diagram” of eigenvalues λ_k or singular values λ_k^1/2 vs. k. The point k*=Sat which this break occurs should not be confused with a “dimension” Oaf the underlying deterministic dynamics.

A Monte-Carlo test can be applied to ascertain the statistical significance of the oscillatory pairs detected by SSA. The entire time series or parts of it that correspond to trends, oscillatory modes or noise can be reconstructed by using linear combinations of the PCs and EOFs, which provide the reconstructed components (RCs) R_k:

$R_K\left(t\right)=\frac{1}{M_t}\sum _{k\in K}\sum _{j=L_t}^{U_t}A_k\left(t-j+1\right)E_k\left(j\right);$

here K is the set of EOFs on which the reconstruction is based. The values of the normalization factor M_t, as well as of the lower and upper bound of summation L_t and U_t, differ between the central part of the time series and the vicinity of its endpoints.

Mufti-channel SSA (or M-SSA) is a natural extension of SSA to an L-channel time series of vectors or maps with N data points {X_l(t): l=1, . . . , L; t=1, . . . , N}. The extended EOF (EEOF) analysis is sometimes assumed to be synonymous with M-SSA. The two methods are both extensions of classical principal component analysis (PCA) but they differ in emphasis: EEOF analysis typically utilizes a number Lot spatial channels much greater than the number M of temporal lags, thus limiting the temporal and spectral information. In M-SSA, on the other hand, one usually chooses (ten M-SSA is applied to a few leading PCs of the spatial data, with M chosen large enough to extract detailed temporal and spectral information from the multivariate time series. To avoid a loss of spectral properties, VARIMAX rotation of the spatio-temporal EOFs (ST-EOFs) of the M-SSA and its variations are sometimes used. Alternatively, a closed matrix formulation of the algorithm for the simultaneous rotation of the EOFs by iterative SVD decompositions has been proposed.

Nonlinear Dimensionality Reduction:

High-dimensional data, meaning data that requires more than two or three dimensions to represent, can be difficult to interpret. One approach to simplification is to assume that the data of interest lie on an embedded non-linear manifold within the higher-dimensional space. If the manifold is of low enough dimension, the data can be visualized in the low-dimensional space. Non-linear methods can be broadly classified into two groups: those that provide a mapping (either from the high-dimensional space to the low-dimensional embedding or vice versa), and those that just give a visualization. In the context of machine learning, mapping methods may be viewed as a preliminary feature extraction step, after which pattern recognition algorithms are applied. Typically, those that just give a visualization are based on proximity data—that is, distance measurements. Related Linear Decomposition Methods include Independent component analysis (ICA), Principal component analysis (PCA) (also called Karhunen-Loève transform—KLT), Singular value decomposition (SVD), and Factor analysis.

The self-organizing map (SOM, also called Kohonen map) and its probabilistic variant generative topographic mapping (GTM) use a point representation in the embedded space to form a latent variable model based on a non-linear mapping from the embedded space to the high-dimensional space. These techniques are related to work on density networks, which also are based around the same probabilistic model.

Principal curves and manifolds give the natural geometric framework for nonlinear dimensionality reduction and extend the geometric interpretation of PCA by explicitly constructing an embedded manifold, and by encoding using standard geometric projection onto the manifold. How to define the “simplicity” of the manifold is problem-dependent. However, it is commonly measured by the intrinsic dimensionality and/or the smoothness of the manifold. Usually, the principal manifold is defined as a solution to an optimization problem. The objective function includes quality of data approximation and some penalty terms for the bending of the manifold. The popular initial approximations are generated by linear PCA, Kohonen's SOM or autoencoders. The elastic map method provides the expectation-maximization algorithm for principal manifold learning with minimization of quadratic energy functional at the “maximization” step.

An autoencoder is a feed-forward neural network which is trained to approximate the identity function. That is, it is trained to map from a vector of values to the same vector. When used for dimensionality reduction purposes, one of the hidden layers in the network is limited to contain only a small number of network units. Thus, the network must learn to encode the vector into a small number of dimensions and then decode it back into the original space. Thus, the first half of the network is a model which maps from high to low-dimensional space, and the second half maps from low to high-dimensional space. Although the idea of autoencoders is quite old, training of deep autoencoders has only recently become possible through the use of restricted Boltzmann machines and stacked denoising autoencoders. Related to autoencoders is the NeuroScale algorithm, which uses stress functions inspired by multidimensional scaling and Sammon mappings (see below) to learn a non-linear mapping from the high-dimensional to the embedded space. The mappings in NeuroScale are based on radial basis function networks.

Gaussian process latent variable models (GPLVM) are probabilistic dimensionality reduction methods that use Gaussian Processes (GPs) to find a lower dimensional non-linear embedding of high dimensional data. They extend the Probabilistic formulation of PCA. The model is defined probabilistically, and the latent variables are then marginalized, and parameters are obtained by maximizing the likelihood. Like kernel PCA, they use a kernel function to form a nonlinear mapping (in the form of a Gaussian process). However, the GPLVM maps from the embedded(latent) space to the data space (like density networks and GTM) whereas in kernel PCA is opposite. It permits visualization of high dimensional data and construction of a shared manifold model between two observation spaces. GPLVM and its many variants have been proposed specially for human motion modeling, e.g., back constrained GPLVM, GP dynamic model (GPDM), balanced GPDM (B-GPDM) and topologically constrained GPDM. To capture the coupling effect of the pose and gait manifolds in the gait analysis, a multi-layer joint gait-pose manifold was proposed.

Curvilinear component analysis (CCA) looks for the configuration of points in the output space that preserves original distances as much as possible while focusing on small distances in the output space (conversely to Sammon's mapping which focuses on small distances in original space). It should be noticed that CCA, as an iterative learning algorithm, actually starts with a focus on large distances (like the Sammon algorithm), then gradually change focus to small distances. The small distance information will overwrite the large distance information if compromises between the two have to be made. The stress function of CCA is related to a sum of right Bregman divergences. Curvilinear distance analysis (CDA) trains a self-organizing neural network to fit the manifold and seeks to preserve geodesic distances in its embedding. It is based on Curvilinear Component Analysis (which extended Sammon's mapping), but uses geodesic distances instead. Diffeomorphic Dimensionality Reduction or Diffeomap learns a smooth diffeomorphic mapping which transports the data onto a lower-dimensional linear subspace. The method solves for a smooth time indexed vector field such that flows along the field which starts at the data points will end at a lower-dimensional linear subspace, thereby attempting to preserve pairwise differences under both the forward and inverse mapping.

Perhaps the most widely used algorithm for manifold learning is Kernel principal component analysis (kernel PCA). It is a combination of PCA and the kernel trick PCA begins by computing the covariance matrix of the M×n Matrix X. It then projects the data onto the first k eigenvectors of that matrix. KPCA begins by computing the covariance matrix of the data after being transformed into a higher-dimensional space. It then projects the transformed data onto the first k eigenvectors of that matrix, just like PCA. It uses the kernel trick to factor away much of the computation, such that the entire process can be performed without actually computing Φ(x). Of course, Φ must be chosen such that it has a known corresponding kernel.

Laplacian Eigenmaps, (also known as Local Linear Eigenmaps, LLE) are special cases of kernel PCA, performed by constructing a data-dependent kernel matrix. KPCA has an internal model, so it can be used to map points onto its embedding that were not available at training time. Laplacian Eigenmaps uses spectral techniques to perform dimensionality reduction. This technique relies on the basic assumption that the data lies in a low-dimensional manifold in a high-dimensional space. This algorithm cannot embed out of sample points, but techniques based on Reproducing kernel Hilbert space regularization exist for adding this capability. Such techniques can be applied to other nonlinear dimensionality reduction algorithms as well. Traditional techniques like principal component analysis do not consider the intrinsic geometry of the data. Laplacian eigenmaps builds a graph from neighborhood information of the data set. Each data point serves as a node on the graph and connectivity between nodes is governed by the proximity of neighboring points (using e.g. the k-nearest neighbor algorithm). The graph thus generated can be considered as a discrete approximation of the low-dimensional manifold in the high-dimensional space. Minimization of a cost function based on the graph ensures that points close to each other on the manifold are mapped close to each other in the low-dimensional space, preserving local distances. The eigenfunctions of the Laplace-Beltrami operator on the manifold serve as the embedding dimensions, since under mild conditions this operator has a countable spectrum that is a basis for square integrable functions on the manifold (compare to Fourier series on the unit circle manifold). Attempts to place Laplacian eigenmaps on solid theoretical ground have met with some success, as under certain nonrestrictive assumptions, the graph Laplacian matrix has been shown to converge to the Laplace-Beltrami operator as the number of points goes to infinity. In classification applications, low dimension manifolds can be used to model data classes which can be defined from sets of observed instances. Each observed instance can be described by two independent factors termed ‘content’ and ‘style’, where ‘content’ is the invariant factor related to the essence of the class and ‘style’ expresses variations in that class between instances. Unfortunately, Laplacian Eigenmaps may fail to produce a coherent representation of a class of interest when training data consist of instances varying significantly in terms of style. In the case of classes which are represented by multivariate sequences, Structural Laplacian Eigenmaps has been proposed to overcome this issue by adding additional constraints within the Laplacian Eigenmaps neighborhood information graph to better reflect the intrinsic structure of the class. More specifically, the graph is used to encode both the sequential structure of the multivariate sequences and, to minimize stylistic variations, the proximity between data points of different sequences or even within a sequence, if it contains repetitions. Using dynamic time warping, proximity is detected by finding correspondences between and within sections of the multivariate sequences that exhibit high similarity. Like LLE, Hessian LLE is also based on sparse matrix techniques. It tends to yield results of a much higher quality than LLE. Unfortunately, it has a very costly computational complexity, so it is not well-suited for heavily sampled manifolds. It has no internal model. Modified LLE (MLLE) is another LLE variant which uses multiple weights in each neighborhood to address the local weight matrix conditioning problem which leads to distortions in LLE maps. MLLE produces robust projections similar to Hessian LLE, but without the significant additional computational cost.

Manifold alignment takes advantage of the assumption that disparate data sets produced by similar generating processes will share a similar underlying manifold representation. By learning projections from each original space to the shared manifold, correspondences are recovered and knowledge from one domain can be transferred to another. Most manifold alignment techniques consider only two data sets, but the concept extends to arbitrarily many initial data sets. Diffusion maps leverage the relationship between heat diffusion and a random walk (Markov Chain); an analogy is drawn between the diffusion operator on a manifold and a Markov transition matrix operating on functions defined on the graph whose nodes were sampled from the manifold. The relational perspective map is a multidimensional scaling algorithm. The algorithm finds a configuration of data points on a manifold by simulating a mufti-particle dynamic system on a closed manifold, where data points are mapped to particles and distances (or dissimilarity) between data points represent a repulsive force. As the manifold gradually grows in size, the multi-particle system cools down gradually and converges to a configuration that reflects the distance information of the data points. Local tangent space alignment (LISA) is based on the intuition that when a manifold is correctly unfolded, all of the tangent hyperplanes to the manifold will become aligned. It begins by computing the k-nearest neighbors of every point. It computes the tangent space at every point by computing the d-first principal components in each local neighborhood. It then optimizes to find an embedding that aligns the tangent spaces. Local Multidimensional Scaling performs multidimensional scaling in local regions and then uses convex optimization to fit all the pieces together.

Maximum Variance Unfolding was formerly known as Semidefinite Embedding. The intuition for this algorithm is that when a manifold is properly unfolded, the variance over the points is maximized. This algorithm also begins by finding the k-nearest neighbors of every point. It then seeks to solve the problem of maximizing the distance between all non-neighboring points, constrained such that the distances between neighboring points are preserved. Nonlinear PCA (NLPCA) uses backpropagation to train a multi-layer perception (MLP) to fit to a manifold. Unlike typical MLP training, which only updates the weights, NLPCA updates both the weights and the inputs. That is, both the weights and inputs are treated as latent values. After training, the latent inputs are a low-dimensional representation of the observed vectors, and the MLP maps from that low-dimensional representation to the high-dimensional observation space. Manifold Sculpting uses graduated optimization to find an embedding. Like other algorithms, it computes the k-nea rest neighbors and tries to seek an embedding that preserves relationships in local neighborhoods. It slowly scales variance out of higher dimensions, while simultaneously adjusting points in lower dimensions to preserve those relationships.

Ruffini (2015) discusses Multichannel transcranial current stimulation (tCS) systems that offer the possibility of EEG-guided optimized, non-invasive brain stimulation. A tCS electric field realistic brain model is used to create a forward “lead-field” matrix and, from that, an EEG inverter is employed for cortical mapping. Starting from EEG, 2D cortical surface dipole fields are defined that could produce the observed EEG electrode voltages. Schestatsky et al. (2017) discuss transcranial direct current stimulation (tDCS), which stimulates through the scalp with a constant electric current that induces shifts in neuronal membrane excitability, resulting in secondary changes in cortical activity. Although tDCS has most of its neuromodulatory effects on the underlying cortex, tDCS effects can also be observed in distant neural networks. Concomitant EEG monitoring of the effects of tDCS can provide valuable information on the mechanisms of tDCS. EEG findings can be an important surrogate marker for the effects of tDCS and thus can be used to optimize its parameters. This combined EEG-tDCS system can also be used for preventive treatment of neurological conditions characterized by abnormal peaks of cortical excitability, such as seizures. Such a system would be the basis of a non-invasive closed-loop device. tDCS and EEG can be used concurrently. See Reference List Table 16.

EEG analysis approaches have emerged, in which event-related changes in EEG dynamics in single event-related data records are analyzed. See Allen D. Malony et al., Computational Neuroinformatics for Integrated Electromagnetic Neuroimaging and Analysis, PAR-99-138. Pfurtscheller, reported a method for quantifying the average transient suppression of alpha band (circa 10-Hz) activity following stimulation. Event-related desynchronization (ERD, spectral amplitude decreases), and event-related synchronization (ERS, spectral amplitude increases) are observed in a variety of narrow frequency bands (4-40 Hz) which are systematically dependent on task and cognitive state variables as well as on stimulus parameters. Ma keig (1993) was reported event-related changes in the full EEG spectrum, yielding a 2-D time/frequency measure he called the event-related spectral perturbation (ERSP). This method avoided problems associated with the analysis of a priori narrow frequency bands, since bands of interest for the analysis could be based on significant features of the complete time/frequency transform. Rappelsburger et al. introduced event-related coherence (ERCOH). A wide variety of other signal processing measures have been tested for use on EEG and/or MEG data, including dimensionality measures based on chaos theory and the bispectrum. Use of neural networks has also been proposed for EEG pattern recognition applied to clinical and practical problems, though usually these methods have not been employed with the aim of explicitly modeling the neurodynamics involved. Neurodynamics is the mobilization of the nervous system as an approach to physical treatment. The method relies on influencing pain and other neural physiology via the mechanical treatment of neural tissues and the non-neural structures surrounding the nervous system. The body presents the nervous system with a mechanical interface via the musculoskeletal system. With movement, the musculoskeletal system exerts non-uniform stresses and movement in neural tissues, depending on the local anatomical and mechanical characteristics and the pattern of body movement. This activates an array of mechanical and physiological responses in neural tissues. These responses include neural sliding, pressurization, elongation, tension, and changes in intraneural microcirculation, axonal transport, and impulse traffic.

The availability of and interest in larger and larger numbers of EEG (and MEG) channels led immediately to the question of how to combine data from different channels. Donchin advocated the use of linear factor analysis methods based on principal component analysis (PCA) for this purpose. Temporal PCA assumes that the time course of activation of each derived component is the same in all data conditions. Because this is unreasonable for many data sets, spatial PCA (usually followed by a component rotation procedure such as Varimax or Promax) is of potentially greater interest. To this end, several variants of PCA have been proposed for ERP decomposition.

Bell and Sejnowski published an iterative algorithm based on information theory for decomposing linearly mixed signals into temporally independent by minimizing their mutual information. First approaches to blind source separation minimized third and fourth-order correlations among the observed variables and achieved limited success in simulations. A generalized approach uses a simple neural network algorithm that used joint information maximization or ‘infomax’ as a training criterion. By using a compressive nonlinearity to transform the data and then following the entropy gradient of the resulting mixtures, ten recorded voice and music sound sources were unmixed. A similar approach was used for performing blind deconvolution, and the ‘infomax’ method was used for decomposition of visual scenes.

Blind decomposition of time series applies to the infomax independent component analysis (ICA) for decomposition of EEG and event-related potential (ERP) data and reported the use of ICA to monitor alertness. This separated artifacts and EEG data into constituent components defined by spatial stability and temporal independence. ICA can also be used to remove artifacts from continuous or event-related (single-trial) EEG data prior to averaging. Viga no et al. (1997).

Since the publication of the original infomax ICA algorithm, several extensions have been proposed. Incorporation of a ‘natural gradient’ term avoided matrix inversions, greatly speeding the convergence of the algorithm and making it practical for use with personal computers on large data EEG and fMRI data sets. An initial ‘sphering’ step further increased the reliability of convergence of the algorithm. The original algorithm assumed that sources have ‘sparse’ (super-Gaussian) distributions of activation values. This restriction has recently been relaxed in an ‘extended-ICA’ algorithm that allows both super-Gaussian and sub-Gaussian sources to be identified. A number of variant ICA algorithms have appeared in the signal processing literature. In general, these make more specific assumptions about the temporal or spatial structure of the components to be separated and typically are more computationally intensive than the infomax algorithm.

Since individual electrodes (or magnetic sensors) each record a mixture of brain and non-brain sources, spectral measures are difficult to interpret and compare across scalp channels. For example, an increase in coherence between two electrode signals may reflect the activation of a strong brain source projecting to both electrodes or the deactivation of a brain generator projecting mainly to one of the electrodes. If independent components of the EEG (or MEG) data can be considered to measure activity within functionally distinct brain networks, however, event-related coherence between independent components may reveal transient, event-related changes in their coupling and decoupling (at one or more EEG/MEG frequencies). ERCOH analysis has been applied to independent EEG components in a selective attention task.

Because sleep patterns have characteristic eye movements, e.g., rapid eye movements (REM), EOG may provide in indication of sleep state. Electromyography: Electromyography (EMG) evaluates and records electrical activity produced by skeletal muscles. EMG reflects muscle tone, which can be indicative of a sleep state. See, Reference List Table 17.

SUMMARY OF THE INVENTION

The present invention provides a system and method for inducing brain states in a subject human or animal which represent or are conducive to sleep according to a natural sleep cycle, based on brain activity patterns (e.g., brainwaves) from a representative of the same species, or the same individual, which are processed and converted to a stimulatory signal that is then imposed on the subject. At any time, the stimulation of the subject seeks to induce EEG patterns that are representative of a target sleep state. The sequence of sleep states may be from the species or individual, or designed according to an algorithm designed or optimized to achieve the overall sleep pattern comprising the desired sequence of sleep states. Advantageously, the subject is monitored during stimulation to measure response to the stimulation, and/or interruption of the target sleep cycle. In case of interruption of the sleep cycle, the stimulation is restarted based on the state of the subject and, therefore, progresses again through a natural sequence of sleep states.

In some cases, the goal is to achieve a sleep cycle which differs from a normal or natural sleep cycle and may be modified or controlled in the case of organic pathology or mental illness. The preferred stimulation is non-invasive, and also preferably not electrical. However, transcranial magnetic stimulation, e.g., subthreshold PEMF, may be employed, alone or with auditory, visual, or other stimulation. For example, while visual stimulation may be contraindicated for entry into respectively deeper sleep states, it may be advantageously used for moving the subject to a shallower sleep state, and to awaken the subject. However, such auditory or visual stimulation preferably is modulated to synchronize or control brainwave patterns, in addition to any overt sensory effects that may be provided. The measurement of brain activity, and brain stimulation may be according to the known methods described hereinabove, without limitation, though the preferred implementation is such that the subject need not be encumbered by bulky, uncomfortable, expensive, or exotic equipment, and can be used unmonitored in a home environment. However, the invention is not limited by such constraints, so long as the sleep pattern is effectively controlled.

In some cases, the invention is used for other than sleep induction, and may also be used to control other stages of consciousness or other mental or emotional states, and preferably a desired sequence of states based on subject biology.

Brain-states, which correlate with specific cognitive states, may be monitored with non-invasive techniques such as EEG and MEG that indirectly measure cortical activity. These cortical signatures provide insight into the neuronal activity, which has been used to identify abnormal cortical function in numerous neurological and psychiatric conditions. Further, the induction of entrained cortical rhythms via transcranial stimulation is effective in imbuing brain-states correlated with such cortical rhythms. See, Poltorak, Alexander. 2019. “On the Possibility of Transplanting Mental States.” OSF Preprints. April 16. doi:10.31219/osf.io/sjqfx.

It has been suggested by Crick and Koch [Crick F, Koch C Towards a neurobiological theory of consciousness. Seminars in the Neurosciences, 1990; (2), 263-75.] (see also [Rees G, Kreiman G, Koch C Neural correlates of consciousness in humans. Nature Reviews. Neuroscience 2002; 3(4)261-270. doi.org/10.1038/nrn783]) that every mental state is expressed through unique neural signals, such as frequency oscillations, that are correlated with mood, cognition and motor functions. It is thus possible to induce a desired mental state by replicating its neural correlates.

Functional neuroimaging, such as electroencephalography (EEG) or magnetoencephalography (MEG), can capture the neuronal activity of localized brain regions which correlate with distinct cognitive or behavioral states (mental states). EEG recordings have demonstrated, for example, that the pattern of brain activity changes during meditative acts, and frontal cortex EEG activity has been associated with emotion induction and regulation Nu et al. [Yu X Fumoto M, Nakatani Y, Sekiya ma T, Kikuchi H, Seki Y, Sato-Suzuki I, Arita H. Activation of the anterior prefrontal cortex and serotonergic system is associated with improvements in mood and EEG changes induced by Zen meditation practice in novices. Intl of Psychophysiology 2011; 80(2),103-111.doi.org/10.1016/j.ijpsycho.2011.02.004]; Dennis and Solomon [Dennis T A, Solomon B. Frontal EEG and emotion regulation: electrocortical activity in response to emotional film dips is associated with reduced mood induction and attention interference effects. Biological Psychology 2010; 85(3),456-464. doi.org/10.1016/j.biopsycho.2010.09.008]). EEG recordings reflect ionic fluctuations resultant of neuronal communication in the cortex arising from dendritic depolarizations (Nunez and Srinivasan [Nunez P L, Srinivasan R. Electric fields of the brain: the neurophysics of EEG (2 ed). Oxford: Oxford Univ. Press; 2006.]). Alternatively, MEG measurements reflect intracellular ionic fluctuations, which similarly result from action potentials (Hämäläinen et al. [Hämäläinen M, Han R, Ilmoniemi R J, Knuutila J, Lounasmaa O V. Magnetoencephalography—theory, instrumentation, and applications to noninvasive studies of the working human brain. Rev. Mod. Phys. 1993; 65, 413. doi.org/10.1103/revmodphys.65.413]). In both cases, the output measures correlate with localized cortical activity.

These EEG or MEG signatures may be inverted in order to stimulate, rather than record, cortical activity. Specifically, transcranial electric stimulation (TES; Annarumma et al. [Annarumma L, D'Atri A, Alfonsi V, De Gennaro L The Efficacy of Transcranial Current Stimulation Techniques to Modulate Resting-State EEG, to Affect Vigilance and to Promote Sleepiness. Brain Sci. 2018; 8(7),137. doi.org/10.3390/brainsci8070137]), including transcranial alternating current stimulation (tACS) and transcranial direct current stimulation (tDCS; Utz et al. [Utz K S, Dimova V, Oppenlander K Kerkhoff G. Electrified minds: transcranial direct current stimulation (tDCS) and galvanic vestibular stimulation (GVS) as methods of non-invasive brain stimulation in neuropsychologya review of current data and future implications. Neuropsychologia 2010; 48(10),2789-2810]) are used to electrically stimulate cortical activity, while transcranial magnetic stimulation (TMS; see [Lawson McLean A. Publication trends in transcranial magnetic stimulation: a 30-year panorama. Brain Stimulation 2019; in press. doi.org/10.1016/j.brs.2019.01.002]) uses a precise magnetic field in order to achieve a similar endpoint of electric current control. Typical brain entrainment methods utilize a constant stimulus (e.g., tDCS) or a synthetic waveform, which may be a step function modulated on a direct current (such as “electrosleep” [Robinovitch L G. Electric analgesia, and electric resuscitation after head failure under chloroform or electrocution. Journal of the American Medical Association 1911; LVI(7),478-481. doi.org/10.1001/jama.1911.02560070010003]), a sinusoid modulated on an oscillatory direct current (osc-tDCS; D'Atri et al. [D'Atri A, DeSimoni E, Gorgoni M, Ferrara M, Ferlazzo F, Rossini P M, De Gennaro L Electrical stimulation of the frontal cortex enhances slow-frequency EEG activity and sleepiness. Neuroscience 2016; 324, 119-130. doi.org/10.1016/j.neuroscience.2016.03.007]), or a fixed frequency modulated on alternating current (tACS; [Rosa, M A, Lisanby, S H. Somatic treatments for mood disorders. Neuropsychopharmacology 2012; 37(1),102-16.10.1038/npp.2011.225]). Helfrich et al., supra, utilized simultaneous tACS stimulation combined with EEG recordings to show that, when tACS was applied to the parieto-occipital lobe of the brain, alpha wave activity increased and became synchronized with the entrainment frequency.

These techniques may be used in order to record and subsequently induce specific brain states. Thus, EEG/MEG may be used to record mental states, which may then be applied via TES (tDCS, osc-tDCS, tACS) or TMS in order to replicate the cognitive-behavioral state of the “donor.” This technique has been previously investigated in the domain of sleep [Gebodh N, Vacchi L, Adair D, UnalG, Poltorak A, Poltora k V, Bikson M. Proceedings #11: Replay of Endogenous Sleep Rhythms to Produce Sleepiness. Brain Stimulation: Basic, Translational, and Clinical Research in Neuromodulation 2019; 12(2), e71-e72. doi.org/10.1016/j.brs.2018.12.180].

The present technology involves, according to one embodiment, the notion of “transplanting” mental states including sleep, attention, and learning, as well as emotional valence. See FIG. 28. Attention-states in the brain are primarily the result of the cognitive process of suppressing the detection of erroneous stimuli. This cognitive state is associated with specific neuronal oscillations (Schroeder et al. [Schroeder C E, Wilson D A Radman T, Scharf man H, Lakatos P. Dynamics of Active Sensing and perceptual selection. Current Opinion in Neurobiology 2010; 20(2),172-176. doi.org/10.1016/j.conb.2010.02.010]) which may be captured via EEG or MEG. The neural oscillations associated with attention have been shown to be disrupted in a number of conditions including epilepsy (Besle et al., [Besle J, Schevon E A Mehta A D, Lakatos P, Goodman R R, McKhann G M, Emerson R G, Schroeder, C E. Tuning of the human neocortex to the temporal dynamics of attended events. The Journal of Neuroscience: The Official Journal of the Society for Neuroscience 2011; 31(9)3176-3185. doi.org/10.1523/JNEUROSCI.4518-10.2011]), dyslexia (Thomson et al. [Thomson J M, Goswami U. Rhythmic processing in children with developmental dyslexia: auditory and motor rhythms link to reading and spelling. Journal of Physiology, Paris 2008; 102(1-3),120-129. doi.org/10.1016/j.jphysparis.2008.03.007]; Leong et al. [Leong V, Goswami U. Assessment of rhythmic entrainment at multiple timescales in dyslexia: evidence for disruption to syllable timing. Hearing Research 2014; 308, 141-161. doi.org/10.1016/j.heares.2013.07.015]; Soltesz et al. [Soltesz F, Szucs D, Leong V, White S, Goswami U. Differential entrainment of neuroelectric delta oscillations in developmental dyslexia. PIoS One 2013; 8(10), e76608. doi.org/10.1371/journal.pone.0076608]), and schizophrenia (Lakatos et al. [Lakatos P, Schroeder C E, Leitman D I, Javilt D C. Predictive suppression of cortical excitability and its deficit in schizophrenia. The Journal of Neuroscience: The Official Journal of the Society for Neuroscience 2013; 33(28),11692-11702. doi.org/10.1523/JNEUROSCI.0010-13.2013]). Therefore, the acquisition of a brainwave signature during states of attention in a healthy “donor” may prove valuable when applied to a recipient exhibiting attention deficits associated with disrupted or otherwise irregular cortical oscillations. Previous research shows that memory functions are acutely sensitive to neural entrainment and may be disrupted via TMS (Hanslmayr et al. [Hanslmayr S, Matusche k J, Fellner M-C Entrainment of Prefrontal Beta Oscillations Induces an Endogenous Echo and Impairs Memory Formation. Current Biology 2014; 24(8),904-909. doi.org/10.1016/j.cub.2014.03.007]) indicating the possibility of an inverse, positive entrainment of these oscillations.

Similarly, emotional arousal and valence are correlated with distinct cortical signatures observable through EEG (Allen et al. [Allen J J B, Keune P M, Schönenberg M, Nusslock R. Frontal EEG alpha asymmetry and emotion: From neural underpinnings and methodological considerations to psychopathology and social cognition. Psychophysiology 2018; 55(1). doi.org/10.1111/psyp.13028]). Previous data indicate that happiness resultant of musical experience, for instance, is associated with increased theta frequency oscillations in the left frontal cortical hemisphere (Rogenmoser et al. [Rogenmosert, Zollinger N, Elmer S, Jancke L Independent component processes underlying emotions during natural music listening. Social Cognitive and Affective Neuroscience 2016; 11(9),1428-1439. doi.org/10.1093/scan/nsw048]). Cortical oscillations associated with negative affect conversely correlate with decreased theta frequency oscillations in this same region. Notably, aberrant cortical oscillations have been observed in a range of affective disorders including major depression (Van der Vinne et al. [Van der Vinne N, Vollebregt M A, van Puffen M J A M, Arns M. Frontal alpha asymmetry as a diagnostic marker in depression: Fact or fiction? A meta-analysis. Neurolmage. Clinical 2017; 16, 79-87. doi.org/10.1016/j.nic1.2017.07.006]). Indeed, the left frontal hemisphere exhibits disrupted cortical rhythms in patients diagnosed with major depression as compared to healthy controls (Nusslock et al. [Nusslock R, Shackman A I, McMenamin B W, Greischar L L, Davidson R J, Kovacs M. Comorbid anxiety moderates the relationship between depression history and prefrontal EEG asymmetry. Psychophysiology 2018; 55(1). doi.org/10.1111/psyp.12953]). Similar data have highlighted cortical asymmetry of frontal lobe oscillations in post-traumatic stress disorder (MD; Meyer et al. [Meyer T, Quaedflieg C W E M, Weijland K Schruers K Merckelbach H, Smeets T. Frontal EEG asymmetry during symptom provocation predicts subjective responses to intrusions in survivors with and without MD. Psychophysiology 2018; 55(1). doi.org/10.1111/psyp.12779]). Simple cortical entrainment via binaural beat stimulation has already proven adequate for inducing specific emotional states (Chaieb et al. [Chaieb L, Wilped E C, Reber T P, Felli Auditory Beat Stimulation and its Effects on Cognition and Mood States. Frontiers in Psychiatry 2015; 6. doi.org/10.3389/fpsyt.2015.00070]). More directly, cranial electrotherapy has been demonstrated as an efficacious treatment for depression, anxiety, and certain forms of insomnia (Kirsch et al. [Kirsch D L, Nichols F. Cranial Electrotherapy Stimulation for Treatment of Anxiety, Depression, and Insomnia. Psychiatric Clinics of North America 2013; 36 (1), 169-176. doi.org/10.1016/j.psc2013.01.006]). Certain forms of depression may respond better to transcranial approaches, such as TMS, as has been demonstrated in early data on patients with treatment-resistant major depression (Rosenberg et al. [Rosenberg O, Shoenfeld N, Zangen A, Kotler M, Dannon P N. Deep TMS in a resistant major depressive disorder a brief report. Depression and Anxiety 2010; 27(5), 465-469. doi.org/10.1002/da.20689]).

This approach to “transplant” (transfer) mental states by replicating neural correlates of the donors state in a recipient is founded on two main principles. First, a large body of literature has identified distinct, measurable cortical signatures associated with specific brain-states ranging from those defining the sleep/wake cycle to those underlying emotional experience. Second, TES and TMS have been repeatedly demonstrated as efficacious, safe means by which cortical rhythms may be entrained with a high degree of location-specificity. Together, these findings provide the basis for the hypothesis that mental states can be “transplanted” (transferred) and provide the means by which a cortical signature may be obtained via EEG or EMG associated with a desired mental state of a “donor” that may, in turn, be processed, inverted, and subsequently applied to a recipient in order to induce said cognitive state via cortical rhythm entrainment using tACS, TMS or other stimuli such as light or sound. Theoretical considerations suggest that this hypothesis is plausible and deserves experimental verification. Importantly, using cortical signatures acquired from a “donor,” rather than a fixed-frequency or synthetic waveform applications as is currently typical for TES techniques, offers the distinct advantage of replicating mufti-phasic, temporally dynamic signals are more likely to induce naturalistic mind-states holistically. This technique may, therefore, provide a novel approach to the non-invasive treatment of a variety of disorders whose current treatments are limited to pharmacotherapeutic interventions.

Sleep disorders affect a significant portion of the adult population. Between 50 and 70 million adults in the U.S. have a sleep disorder. (Ohayon M M. Epidemiology of insomnia: what we know and what we still need to learn. Sleep medicine reviews. 2002; 6(2):97-111.) Insomnia is the most common specific sleep disorder, with short-term issues reported by about 30% of adults and chronic insomnia by 10%. (Kessler R C, Berglund P A Coulouvrat C, et al. Insomnia and the performance of US workers: results from the America insomnia survey. Sleep. 2011; 34(9)1161-1171; Sateia M I, Doghramji K Hauri P I, Morin C M. Evaluation of chronic insomnia. An American Academy of Sleep Medicine review. Sleep. 2000; 23(2):243-308.) Chronic insomnia is associated with deterioration of memory, adverse effects on endocrine functions and immune responses, and an increase in the risk of obesity and diabetes Sateia et al. 2000; Taylor D J, Mallory U, Lichstein K L, Durrence H H, Riedel B W, Bush A I. Comorbidity of chronic insomnia with medical problems. Sleep. 2007; 30(2):213-218). While at any age, managing insomnia is a challenge, it is especially a critical condition in the elderly due to age-related increases in comorbid medical conditions and medication use, as well as age-related changes in sleep structure, which shorten sleep time and impair sleep quality. (Ancoli-Israel S. Insomnia in the elderly: a review for the primary care practitioner. Sleep. 2000; 23:523-30; discussion 536-28; Buysse D J. Insomnia, depression, and aging. Assessing sleep and mood interactions in older adults. Geriatrics (Basel, Switzerland). 2004; 59(2):47-51; quiz 52.) As a result, decreased sleep quality is one of the most common health complaints of older adults. Medications are widely prescribed for relief from insomnia. However, sleep-promoting agents, such as hypnotic drugs, can produce adverse effects, particularly in the elderly. (Sateia M J, Buysse D J, Krystal A D, Neubauer D N. Adverse Effects of Hypnotic Medications. J Clin Sleep Med. Jun. 15, 2017; 13(6):839.) Even natural supplements, such as melatonin, can cause some side effects, including headache, depression, daytime sleepiness, dizziness, stomach cramps, and irritability. (Buscemi N, Vandermeer B, Hooton N, et al. The efficacy and safety of exogenous melatonin for primary sleep disorders a meta-analysis. Journal of general internal medicine. 2005; 20(12):1151-1158.)

Aside from the general deterioration of sleep quality with age in adult population, the deterioration in quantity and quality of the slow-wave sleep (SWS), which is non-REM deep sleep, is particularly troubling. (Roth T. Slow wave sleep: does it matter?Journal of clinical sleep medicine: JCSM: official publication of the American Academy of Sleep Medicine. 2009; 5(2 Suppl):54). SINS plays an important role in cerebral restoration and recovery in humans. Studies have shown that a 15% reduction in the amounts of SWS and increased number and duration of awakenings are associated with normal aging. (Chinoy E D, Frey D J, Kaslovsky D N, Meyer F G, Wrightlr K P. Age-related changes in slow wave activity rise time and NREM sleep EEG with and without zolpidem in healthy young and older adults. Sleep medicine. 2014; 15(9):1037-1045.) Experimental disruption of SWS have been shown to increase shallow sleep, sleep fragmentation, daytime sleep propensity, and impair daytime function. (Dijk DJ. Regulation and functional correlates of slow wave sleep. J Clin Sleep Med. Apr. 15 2009; 5(2 Suppl):56-15; Restoring Deep, Slow Wave Sleep to Enhance Health and Increase Lifespan, nutritionreview.org/2014/07/restoring-slow-wave-sleep-shown-enhance-health-increase-lifespan/(2014)). Given that SWS contributes to sleep continuity, enhancement of SWS may lead to improvements in sleep quality and daytime function in patients with insomnia and the elderly. Furthermore, accumulating evidence point to the SWS is the time when short-term memory is consolidated into long-term memory. (Born J. Slow-wave sleep and the consolidation of long-term memory. The World Journal of Biological Psychiatry. 2010; 11(supl):16-21.) Recent research connects the deterioration of the SWS with early onset of Alzheimer's disease and other forms of dementia. (Petit D, Gagnon J-F, Fantini M L, Ferini-Strambi Montplaisir J. Sleep and quantitative EEG in neurodegenerative disorders. Journal of psychosomatic research. 2004; 56(5):487-496; McCurry S M, Ancoli-Israel S. Sleep dysfunction in Alzheimer's disease and other dementias. Current treatment options in neurology. 2003; 5(3):261-272). It is also suggested that the loss of SWS stage may play a role in these debilitating age-related diseases. (Mattis J, Sehgal A. Circadian rhythms, sleep, and disorders of aging. Trends in Endocrinology Er Metabolism. 2016; 27(4):192-203). Unfortunately, most standard sleeping pills, while alleviating insomnia, do little to improve the SWS. (Walsh R Enhancement of slow wave sleep: implications for insomnia. Journal of clinical sleep medicine: JCSM: official publication of the American Academy of Sleep Medicine. 2009; 5(2 Suppl):527.) Some evidence suggests that some hypnotic drugs change the structure of sleep, adversely affecting the SWS (Sateia et al. (2017); Walsh (2009). Hence, there is an unmet need for non-pharmacological techniques for promoting sleep, particularly, the deep non-REM sleep stage (SWS) lacking in the elderly population.

One of the promising non-pharmacological approaches to promoting sleep is neuromodulation via light, sound, and/or transcranial electric Stimulation (TES). Limited human trials conducted by Neuroenhancement Lab in collaboration with the Neuromodulation Laboratory at The City College of New York (CUNY) showed promise in replicating the desired sleep stage 0 a healthy donor in other subjects (recipients). Electroencephalogram (EEG) of healthy volunteers were recorded as they dozed off entering stage 1 of sleep, as evidenced by the predominance of alpha waves. These EEG recordings were subsequently filtered from noise, inverted, and used for transcranial Endogenous Sleep-Derived stimulation (tESD). Volunteer subjects stimulated with tESD modulated with the indigenous brainwaves recorded in a sleeping donor, quickly dozed off and entered stage 1 of sleep, as evidenced by EEG, heart rate, respiration rate, and post-sleep cognitive test. These results were better as compared to the control arms of the study that included sham stimulation, tDCS, and tACS (10 Hz). These preliminary results suggest that tACS modulated with indigenous brainwaves recorded from a healthy sleeping donor can be used to replicate the desired sleep stage of a healthy donor in another subject.

There is significant research to identify markers of different phases of healthy or pathological sleep; the markers allow to classify observed EEG to one of the phases of sleep/wake categories. The applicants are not aware of any research that aimed at comprehensive identification of all independent components of EEG signals during sleep; and comprehensive analysis of statistically significant inter-dependence of a presence of an independent component with the particular stage of sleep. Comprehensive identification and analysis of independent components associated with sleep would allow to use those components and/or derived signals for a tACS protocol

EEG recordings of brainwaves are obtained and pre-processed from healthy human subjects during various stages of sleep. EEG recordings of three stages of sleep, and while being awake from at least ten healthy subjects (e.g., through public EEG database), which are then smoothed and filtered. The EEG recordings are analyzed to identify statistically significant waveform components correlated with specific sleep stages. A model (e.g., a linear multivariate model) is developed for the coefficients of the components of the EEG, based on sleep stage/wakefulness status; and the statistical significance of the model is measured. Stimulation protocols are developed that can provide safe and effective neurostimulation to induce desired sleep stage.

Great economic burden and the societal cost is incurred due to sleeping disorders, particularly insomnia. Sleep disturbances are common symptoms in adults and are related to various factors, including the use of caffeine, tobacco, and alcohol; sleep habits; and comorbid diseases/.

Epidemiologic studies indicate sleep disorders are affecting a significant portion of adult population. Between 50 and 70 million adults in the U.S. have a sleep disorder. Insomnia is the most common specific sleep disorder, with short-term issues reported by about 30% of adults and chronic insomnia by 10%. (Kessler et al. (2011), Sateia et al. (2000), Ancoli-Israel et al. (2000), Ancoli-Israel S, Roth T. Characteristics of insomnia in the United States: results of the 1991 National Sleep Foundation Survey. I. Sleep. 1999; 22:5347-353). Chronic insomnia is associated with deterioration of memory, adverse effects on endocrine functions and immune responses, and an increase in the risk of obesity and diabetes. (Sateia et al. (2000)). In addition, there is a significant economic burden and societal cost associated with insomnia due to the impact on health care utilization, impact in the work domain, and quality of life. Recent estimates of direct and indirect costs are upwards of 100 billion dollars annually in the United States. (Fullerton DS. The economic impact of insomnia in managed care: a dearer picture emerges. Am J Manag Care. May 2006; 12(8 Suppl):S246-252.) While at any age, managing insomnia is a challenge, it is especially a critical condition in the elderly due to age-related increases in comorbid medical conditions and medication use, as well as age-related changes in sleep structure, which shorten sleep time and impair sleep quality. (Ancoli-Israel (2000). As a result, decreased subjective sleep quality is one of the most common health complaints of older adults.

There is a deterioration of the slow-wave sleep (SWS) in the elderly. Aside from the general deterioration of sleep quality with age in the adult population, the deterioration in quantity and quality of the slow-wave sleep (SWS), which is the deep non-REM sleep, is particularly troubling. (Roth (2009)). SWS plays an important role in cerebral restoration and recovery in humans. It is the most prominent EEG event during sleep and appears as spontaneous large oscillations of the EEG signal occurring approximately once every second in the deepest stage of non-REM sleep. (Achermann P, Dijk D-J, Brunner D P, Borbèly A A. A model of human sleep homeostasis based on EEG slow-wave activity: quantitative comparison of data and simulations. Brain research bulletin. 1993; 31(1-2):97-113.) Studies have shown that a significant decrease (˜15% reduction) in the amounts of SWS and increased number and duration of awakenings are associated with normal aging. (Chinoy et al. (2014)). Given that SWS contributes to sleep continuity and experimental disruption of SWS increases shallow sleep and sleep fragmentation, enhances daytime sleep propensity, and impairs daytime function. (Dijk (2009); NutritionReview.org (2014)), enhancement of SWS may lead to improvements in sleep maintenance and daytime function in patients with insomnia and in the elderly. Furthermore, accumulating evidence point to the SWS as the time when short-term memory is consolidated into long-term memory. (Born (2010)). Recent research connects the deterioration of the SWS with early onset of AlzheimeCs disease and other forms of dementia. (Petit et al. (2010); McCurry et al. (2003)). It is also suggested that the loss of SWS stage may be the culprit for these debilitating age-related diseases. (Mattis et al. (2016)).

SWS enhancement is a potential non-pharmacological therapy for the elderly. Given the pivotal role of slow waves during sleep, it is not surprising that several efforts have been made to increase sleep efficacy by potentiating SWS. Recently, a number of drugs have been shown to increase SWS. Although acting on different synaptic sites, overall the slow wave enhancing the effect of these drugs is mediated by enhancing GABAergic transmission. Specifically, clinical investigations showed that both tiagabine and gaboxadol increased the duration of SWS after sleep restriction. (Walsh (2009); Mathias S, Wetter T C, Steiger A, Lancet M. The GABA uptake inhibitor tiagabine promotes slow wave sleep in normal elderly subjects. Neurobiology of aging. 2001; 22(2):247-253; Walsh J K, Snyder E, Hall J, et al. Slow wave sleep enhancement with gaboxadol reduces daytime sleepiness during sleep restriction. Sleep. May 2008; 31(5):659-672; Feld G B, Wilhelm I, Ma Y, et al. Slow wave sleep induced by GABA agonist tiagabine fails to benefit memory consolidation. Sleep. Sep. 1, 2013; 36(9):1317-1326). Tiagabine also improved performance on cognitive tasks evaluating executive functions and reduced the negative effects of sleep restriction on alertness. (Walsh J K, Randazzo A C, Stone K et al. Tiagabine is associated with sustained attention during sleep restriction: evidence for the value of slow-wave sleep enhancement? Sleep-New York Then Westchester-. 2006; 29(4):433). Although these results are positive, pharmacological approaches to sleep enhancement often raise issues related to dependence and tolerance and are commonly associated with residual daytime side effects. Some evidence suggests that some hypnotic drugs, while alleviating insomnia, change the structure of sleep adversely affecting the SWS. (Sateia (2000); Walsh (2009)). Even natural supplements, such as melatonin, can cause some side effects, including headache, short-term feelings of depression, daytime sleepiness, dizziness, stomach cramps, and irritability. (Buscemi et al. (2005)). Hence, there is an unmet need for a non-pharmacological technique for promoting sleep, particularly in the deep non-REM sleep stage lacking in the elderly population.

Brainwaves, e.g., EEG signals, may be acquired in various ways. Traditional signal acquisition by neurologists and encephalography/EEG technicians involves pasted-on electrodes or caps with arrays of electrodes, e.g., 20-256 electrodes positioned on the scalp. However, in some cases, especially where high spatial resolution is not required, and dominant brainwave patterns are sought, simpler and less controlled EEG acquisition systems may be employed, including through commercially available device intended to interface with smartphones. See, kokoon.io, www.thinkmindsetcom/; www.choosemuse.com (Muse, Muse2); Neurosky; getvi.com Sense); Strickland, Eliza, “In-Ear EEG Makes Unobtrusive Brain-Hacking Gadgets a Real Possibility”, IEEE Spectrum Jul. 7, 2016; Strickland, Eliza, “Wireless Earbuds Will Record Your EEG, Send Brainwave Data To Your Phone”, IEEE Spectrum May 17, 2016. The Unicorn “Hybrid Black” wearable EEG headset provides a headset with eight electrode channels and digital data acquisition electronics (24 bit 250 Hz), intended to provide a brain-computer interface for a distic, control and other tasks. See, www.unicorn-bi.com/. Starkey Laboratories, Inc. US 20190166434 discloses an ear-worn electronic device having a plurality of sensors for EEG signals from a wearer's ear, as a brain-computer interface. A number of designs provide in-ear headphones which integrate EEG electrodes that pick up signals from the ear canal. See Reference List Table 18.

The brain activity of a first subject (a “donor” who is in the desired sleeping state) may be captured by recording neural correlates of the sleep, as expressed by brain activity patterns, such as EEG signals. The representations of the neural correlates of the first subject are used to control stimulation of a second subject (a “recipient”), seeking to induce the same brain activity patterns of the donor in the recipient to assist the recipient in attaining the desired sleep state that had been attained by the donor.

One strategy to enhance deep sleep non-pharmacologically is to stimulate the brain with light, sound, electrical currents, or magnetic fields based on artificial and synthetic stimulation paradigms. Intermittent transcranial direct-current stimulation (tDCS) applied at 0.75 Hz for 5-min intervals separated by 1-min off periods after SWS onset can increase the EEG power in the slow oscillation band (<1 Hz) during the stimulation-free intervals. (Lang N, Siebner H R, Ward N S, et al. How does transcranial DC stimulation of the primary motor cortex alter regional neuronal activity in the human brain? European Journal of Neuroscience. 2005; 22(2):495-504; Marshall L, Helgadoltir H, Molle M, Born J. Boosting slow oscillations during sleep potentiates memory. Nature. Nov. 30 2006; 444(7119):610-613). Similarly, stimulated by tDCS at the beginning of SWS accelerate the SWA homeostatic decay in subjects. (Reato D, Gasca F, Datta A, Bikson M, Marshall L, Parra L C. Transcranial electrical stimulation accelerates human sleep homeostasis. PLoS Comput Biol. 2013; 9(2):e1002898). Furthermore, slow waves can be triggered by directly perturbing the cortex during non-REM sleep using transcranial magnetic stimulation (TMS). (Massimini M, Ferrarelli F, Esser S K, et al. Triggering sleep slow waves by transcranial magnetic stimulation. Proc Natl Acad Sci USA. May 15 2007; 104(20):8496-8501). Other research has focused on the possibility of inducing slow waves in a more physiological natural manner. In a larger study in healthy adults, bilateral electrical stimulation of the vestibular apparatus shortened sleep onset latency in comparison to sham nights where no stimulation was provided. (Krystal A D, Zammit G K Wyaff J K, et al. The effect of vestibular stimulation in a four-hour sleep phase advance model of transient insomnia. J Clin Sleep Med. Aug. 15 2010; 6(4)315-321). The effect of somatosensory and auditory stimulation was also assessed (Krystal et al. 2010; Ngo H V, Ma rtinetz T, Born J, Molle M. Auditory closed-loop stimulation of the sleep slow oscillation enhances memory. Neuron. May 8 2013; 78(3):545-553). While the change observed with somatosensory stimulation was minor, acoustic stimulation was particularly efficacious in enhancing sleep slow waves. Specifically, using an intermittent stimulation, in which tones were played in blocks of 15 s spaced out by stimulation-free intervals, slow waves appeared remarkably large and numerous during the stimulation blocks. (Tononi G, Riedner B, Hulse B, Ferrarelli F, Sarasso S. Enhancing sleep slow waves with natural stimuli. Medicamundi. 2010; 54(2):73-79; Bellesi M, Riedner B A Garcia-Molina G N, Cirelli C, Tononi G. Enhancement of sleep slow waves: underlying mechanisms and practical consequences. Frontiers in systems neuroscience. 2014; 8:208). In addition, high-density EEG studies (hdEEG, 256 channels) showed that the morphology, topography, and traveling patterns of induced slow waves were indistinguishable from those of spontaneous slow waves observed during natural sleep. A recent study found that EEG SWA increased following tone presentation during non-REM sleep (Arzi A, Shedlesky L, Ben-Shaul M, et al. Humans can learn new information during sleep. Nature neuroscience. 2012; 15(10):1460), and slow oscillation activity (0.5-1 Hz) was increased in response to continuous acoustic stimulation at 0.8 Hz starting 2 min before lights were turned off and lasting for 90 min. (Ngo H V, Claussen J C, Born J, Molle M. Induction of slow oscillations by rhythmic acoustic stimulation. J Sleep Res. February 2013; 22(1):22-31). Unlike the previous neurostimulation methods with a dificialand synthetic stimulation paradigms, the present stimulation protocol uses source-derived waveforms, extracted from the indigenous brain activity EEG recordings of the healthy subjects, processed by statistical methods (e.g., principal component analysis, independent component analysis (Ungureanu M, Bigan C, Strungaru R, Lazarescu V. Independent component analysis applied in biomedical signal processing. Measurement Science Review. 2004; 4(2):18) or spatial principal component analysis, autocorrelation, etc.), which separates components of brain activity. These separated brain EEG activities are then modified or modulated and subsequently inverted and used for transcranial Endogenous Sleep-Derived stimulation (tESD). The application of endogenous brain waveform should not only retain the efficacy in triggering SINS but also alleviate the safety concerns that are associated with long-term brain stimulation using synthetic paradigms.

The present technology provides a method of improving sleep by transplanting sleep states—one desired sleep stage, or the sequences of sleep stages—from the first subject (donor) (or from a plurality of donors) to a second subject (recipient). (In some embodiments, the first and the second subject may be the same subject at different points in time, or based on a protocol or algorithm.)

The process seeks to achieve, in the subject, a brainwave pattern, which is derived from a human. The brainwave pattern is complex, representing a superposition of modulated waveforms. The modulation preferably is determined based on brain wave patterns of another subject or plurality of subjects.

Sleep is a natural periodic suspension of consciousness, basically a process that can hardly be influenced in its individual stages by the person sleeping. It is a subconscious (in a technical sense) mental state, representing a resting state, activity pattern, activity rhythm, readiness, receptivity, or other state, often independent of particular inputs. In essence, a sleep state in a particular sleep stage or a sequence of different sleep stages of the first subject (a “donor” who is in a desired sleep stage or goes through a sequence with its individual stages) is captured by recording neural correlates of the sleep state, e.g., as expressed by brain activity patterns, such as EEG or MEG signals. The neural correlates of the first subject, either as direct or recorded representations, may then be used to control a stimulation of the second subject (a “recipient”), seeking to induce the same brain activity patterns in the second subject (recipient) as were present in the first subject (donor), thereby transplanting the sleep state of the first subject (donor), to assist the second subject (recipient) to attain the desired sleep stage that had been attained by the donor. In an alternative embodiment, the signals from the first subject (donor) being in a first sleep stage are employed to prevent the second subject (recipient) from achieving a second sleep stage, wherein the second sleep stage is an undesirable one. Furthermore, the duration and timing of different sleep stages can be controlled in the second subject. This could enable the change of the individual duration or intensity of each sleep stage and the order in which they appear. In some embodiments, the signals from the first subject can be used to trigger sleep in the second subject or to prevent sleep or sleepiness and associated symptoms such as fatigue, lack of concentration, etc.

In one embodiment, brain activity patterns are recorded during a complete sleep cycle or during several such cycles over the course of a normal night sleeping. In some embodiments, the acquiring of the sleep state information is preceded by or followed by identifying the sleep stage, by direct reporting by the first subject (donor) or a n observer, or by automated analysis of the physiological parameters (e.g., brain activity patterns, heartbeat, breathing pattern, oxygen saturation in blood, temperature, eye movement, skin impedance, etc.) or both. In other embodiments, the processing of the brain activity patterns does not seek to classify or characterize it, but rather to filter and transform the information to a form suitable for control of the stimulation of the second subject. In particular, according to this embodiment, the subtleties that are not yet reliably classified in traditional brain activity pattern analysis are respected. For example, it is understood that all brain activity is reflected in synaptic currents and other neural modulation and, therefore, theoretically, conscious and subconscious information is, in theory, accessible through brain activity pattern analysis. Since the available processing technology generally fails to distinguish a large number of different brain activity patterns, that available processing technology, is necessarily deficient, but improving. However, just because a computational algorithm is unavailable to extract the information, does not mean that the information is absent. Therefore, this embodiment employs relatively raw brain activity pattern data, such as filtered or unfiltered EEGs, to control the stimulation of the second subject, without a full comprehension or understanding of exactly what information of significance is present. In one embodiment, brainwaves are recorded and “played back” to another subject, similar to recording and playing back music. Such recording-playback may be digital or analog. Typically, the stimulation may include a low dimensionality stimulus, such as stereo-optic binaural, isotonic tones, tactile, or other sensory stimulation, operating bilaterally, and with control over frequency and phase and/or waveform and/or transcranial stimulation such as TES, tDCS, HD-tDCS, tACS, or TMS. A plurality of different types of stimulation may be applied concurrently, e.g., visual, auditory, other sensory, magnetic, electrical.

Likewise, a present lack of complete understanding of the essential characteristics of the signal components in the brain activity patterns does not prevent their acquisition, storage, communication, and processing (to some extent). The stimulation may be direct, i.e., a visual, auditory, or tactile stimulus corresponding to the brain activity pattern or a derivative or feedback control based on the second subjects brain activity pattern.

To address the foregoing problems, in whole or in part, and/or other problems that may have been observed by persons skilled in the art, the present disclosure provides methods, processes, systems, apparatus, instruments, and/or devices, as described by way of example in implementations set forth below.

While mental states are typically considered internal to the individual, and subjective such states are common across individuals and have determinable physiological and electrophysiological population characteristics. Further, mental states may be externally changed or induced in a manner that bypasses the normal cognitive processes. In some cases, the triggers for the mental state are subjective, and therefore the particular subject-dependent sensory or excitation scheme required to induce a particular state will differ. For example, olfactory stimulation can have different effects on different people, based on differences in the history of exposure, social and cultural norms, and the like. On the other hand, some mental state response triggers are normative, for example, “tearjerker” media.

Mental states are represented in brainwave patterns, and in normal humans, the brainwave patterns and metabolic (e.g. blood flow, omen consumption, etc.) follow prototypical patterns. Therefore, by monitoring brainwave patterns in an individual, a state or series of mental states in that person may be determined or estimated. However, the brainwave patterns may be interrelated with context, other activity, and past history. Further, while prototypical patterns may be observed, there are also individual variations in the patterns. The brainwave patterns may include characteristic spatial and temporal patterns indicative of mental state. The brainwave signals of a person may be processed to extract these patterns, which, for example, may be represented as hemispheric signals within a frequency range of 3-100 Hz. These signals may then be synthesized or modulated into one or more stimulation signals, which are then employed to induce a corresponding mental state into a recipient, in a manner seeking to achieve a similar brainwave pattern from the source. The brainwave pattern to be introduced need not be newly acquired for each case. Rather, signals may be acquired from one or more individuals, to obtain an exemplar for various respective mental state. Once determined, the processed signal representation may be stored in non-volatile memory for later use. However, in cases of complex interaction between a mental state and a context or content or activity, it may be appropriate to derive the signals from a single individual whose context or content-environment or activity is appropriate for the circumstances. Further, in some cases, a single mental state, emotion or mood is not described or fully characterized, and therefore acquiring signals from a source is an efficient exercise.

With a library of target brainwave patterns, a system and method is provided in which a target subject may be immersed in a presentation, which includes not only multimedia content but also a series of defined mental states, emotional states or moods that accompany the multimedia content. In this way, the multimedia presentation becomes fully immersive. The stimulus, in this case, may be provided through a headset, such as a virtual reality or augmented reality headset. This headset is provided with a stereoscopic display, binaural audio, and a set of EEG stimulatory electrodes. These electrodes (if provided) typically deliver a subthreshold signal, which is not painful, which is typically an AC signal which corresponds to the desired frequency, phase, and spatial location of the desired target pattern. The electrodes may also be used to counteract undesired signals, by destructively interfering with them while concurrently imposing the desired patterns. The headset may also generate visual and/or auditory signals which correspond to the desired state. For example, the auditory signals may induce binaural beats, which cause brainwave entrainment. The visual signals may include intensity fluctuations or other modulation patterns, especially those which are subliminal, that are also adapted to cause brainwave entrainment or induction of the desired brainwave pattern.

The headset preferably includes EEG electrodes for receiving feedback from the user. That is, the stimulatory system seeks to achieve a mental state, emotion or mood response from the user. The EEG electrodes permit determination of whether that state is achieved, and if not, what the current state is. It may be that achieving the desired brainwave pattern is state dependent, and therefore that characteristics of the stimulus to achieve the desired state depend on the starting state of the subject. Other ways of determining mental state, emotion, or mood include analysis of facial expression, electromyography (EMG) analysis of facial muscles, explicit user feedback, etc.

An authoring system is provided which permits a content designer to determine what mental states are desired, and then encode those states into media, which is then interpreted by a media reproduction system in order to generate appropriate stimuli. As noted above, the stimuli may be audio, visual, multimedia, other senses, or electrical or magnetic brain stimulation, and therefore a VR headset with transcranial electrical or magnetic stimulation is not required. Further, in some embodiments, the patterns may be directly encoded into the audiovisual content, subliminally encoded. In some cases, the target mental state may be derived from an expert, actor or professional exemplar. The states may be read based on facial expressions, EMG, EEG, or other means, from the actor or exemplar. For example, a prototype exemplar engages in an activity that triggers a response, such as viewing the Grand Canyon or a rhNorks within the Louvre. The responses of the exemplar are then recorded or represented, and preferably brainwave patterns recorded that represent the responses. A representation of the same experience is then presented to the target, with a goal of the target also experiencing the same experience as the exemplar. This is typically a voluntary and disclosed process, so the target will seek to willingly comply with the desired experiences. In some cases, the use of the technology is not disclosed to the target, for example in advertising presentations or billboards. In order for an actor to serve as the exemplar, the emotions achieved by that person must be authentic. However, so-called “method actors” do authentically achieve the emotions they convey. However, in some cases, for example, where facial expressions are used as the indicator of mental state, an actor can present desired facial expressions with inauthentic mental states. The act of making a face corresponding to an emotion often achieves the targeted mental state.

In general, the present technology is directed toward inducing sleep. Note that certain kinds of content are known to assist in induction of sleep, e.g., a lulla bye. However, the present technology encompasses both human comprehensible signals and incomprehensible (noise-like) signals.

In order to calibrate the system, the brain pattern of a person may be measured while in the desired state. The brain patterns acquired for calibration or feedback need not be of the same quality, or precision, or data depth, and indeed may represent responses rather than primary indicia. That is, there may be some asymmetry in the system, between the brainwave patterns representative of a mental state, and the stimulus patterns appropriate for inducing the brain state.

The present invention generally relates to achieving a mental state in a subject by conveying to the brain of the subject patterns of brainwaves. These brainwaves may be artificial or synthetic, or derived from the brain of a second subject (e.g., a person experiencing an authentic experience or engaged in an activity). Typically, the wave patterns of the second subject are derived while the second subject is experiencing an authentic experience.

A special case is where the first and second subjects are the same individual. For example, brainwave patterns are recorded while a subject is in a particular mental state. That same pattern may assist in achieving the same mental state at another time. Thus, there may be a time delay between the acquisition of the brainwave information from the second subject, and exposing the first subject to corresponding stimulation. The signals may be recorded and transmitted.

The temporal pattern may be conveyed or induced non-invasively via light (visible or infrared), sound (or infrasound). Alternately, non-sensory stimulation may be employed, e.g., transcranial direct or alternating current stimulation (tDCS or tACS), transcranial magnetic stimulation (TMS), Deep transcranial magnetic stimulation (Deep TMS, or dTMS), Repetitive Transcranial Magnetic Stimulation (rTIMS) olfactory stimulation, tactile stimulation, or any other means capable of conveying frequency patterns. In a preferred embodiment, normal human senses are employed to stimulate the subject, such as light, sound, smell, and touch. Combinations of stimuli may be employed. In some cases, the stimulus or combination is innate, and therefore largely pan-subject. In other cases, response to a context is learned, and therefore subject-specific. Therefore, feedback from the subject may be appropriate to determine the triggers and stimuli appropriate to achieve a mental state.

This technology may be advantageously used to enhance mental response to a stimulus or context. Still another aspect provides for a change in the mental state. The technology may be used in humans or animals.

The present technology may employ an event-correlated EEG time and/or frequency analysis performed on neuronal activity patterns. In a time-analysis, the signal is analyzed temporally and spatially, generally looking for changes with respect to time and space. In a frequency analysis, over an epoch of analysis, the data, which is typically a time-sequence of samples, is transformed, using e.g., a Fourier transform (FT, or one implementation, the Fast Fourier Transform, FFT), into a frequency domain representation, and the frequencies present during the epoch are analyzed. The window of analysis may be rolling, and so the frequency analysis may be continuous. In a hybrid time-frequency analysis, for example, a wavelet analysis, the data during the epoch is transformed using a “wavelet transform”, e.g., the Discrete Wavelet Transform (DWT) or continuous wavelet transform (CWT), which has the ability to construct a time-frequency representation of a signal that offers very good time and frequency localization. Changes in transformed data over time and space may be analyzed. In general, the spatial aspect of the brainwave analysis is anatomically modeled. In most cases, anatomy is considered universal, but in some cases, there are significant differences. For example, brain injury, psychiatric disease, age, race, native language, training, sex, handedness, and other factors may lead to distinct spatial arrangement of brain function, and therefore when transferring mood from one individual to another, it is preferred to normalize the brain anatomy of both individuals by experiencing roughly the same experiences, and measuring spatial parameters of the EEG or MEG. Note that spatial organization of the brain is highly persistent, absent injury or disease, and therefore, this need only be performed infrequently. However, since electrode placement may be inexact, a spatial calibration may be performed after electrode placement.

Different aspects of EEG magnitude and phase relationships may be captured, to reveal details of the neuronal activity. The “time-frequency analysis” reveals the brain's parallel processing of information, with oscillations at various frequencies within various regions of the brain reflecting multiple neural processes co-occurring and interacting. See, Lis man J, Buzsaki G. A neural coding scheme formed by the combined function of gamma and theta oscillations. Schizophr Bull. Jun. 16, 2008; doi:10.1093/schbul/sbn060. Such a time-frequency analysis may take the form of a wavelet transform analysis. This may be used to assist in integrative and dynamically adaptive information processing. Of course, the transform may be essentially lossless and may be performed in any convenient information domain representation. These EEG-based data analyses reveal the frequency-specific neuronal oscillations and their synchronization in brain functions ranging from sensory processing to higher-order cognition. Therefore, these patterns may be selectively analyzed, for transfer to or induction in, a subject.

A statistical clustering analysis may be performed in high dimension space to isolate or segment regions which act as signal sources, and to characterize the coupling between various regions. This analysis may also be used to establish signal types within each brain region and decision boundaries characterizing transitions between different signal types. These transitions may be state dependent, and therefore the transitions may be detected based on a temporal analysis, rather than merely a concurrent oscillator state.

The various measures make use of the magnitude and/or phase angle information derived from the complex data extracted from the EEG during spectral decomposition and/or temporal/spatial/spectral analysis. Some measures estimate the magnitude or phase consistency of the EEG within one channel across trials, whereas others estimate the consistency of the magnitude or phase differences between channels across trials. Beyond these two families of calculations, there are also measures that examine the coupling between frequencies, within trials and recording sites. Of course, in the realm of time-frequency analysis, many types of relationships can be examined beyond those already mentioned.

These sensory processing specific neuronal oscillations, e.g., brainwave patterns, e.g., of a subject (a “source”) or to a person trained (for example, an actor trained in “the method”) to create the desired state, and can be stored on a tangible medium and/or can be simultaneously conveyed to a recipient making use of the brain's frequency following response nature. See, Galbraith, Gary C, Darlene M. Off ma n, and Todd M. Huffman. “Selective attention affects human brain stem frequency-following response.” Neuroreport 14, no. 5 (2003): 735-738, journals.lww.com/neuroreport/Abstract/2003/04150/Selective_affention affects_human_brain_stem.15.aspx.

According to one embodiment, the stimulation of the second subject is combined with a feedback process, to verify that the second subject has appropriately responded to the stimulation, e.g., has a predefined similarity to the sleep stage as the first subject, has a sleep stage with a predefined difference from the first subject, or has the desired change from a baseline sleep stage. The feedback may be based on brain activity per se, or neural correlates of sleep stage or, alternatively, or, in addition, physical, psychological, or behavioral effects that may be measured, reported or observed. The feedback typically is provided to a controller with at least partial model basis, for the stimulator, which alters stimulation parameters to optimize the stimulation.

As discussed above, the model is typically difficult to define. Therefore, the model-based controller is incompletely defined, and the existence of errors and artifacts is to be expected. However, by employing a model-based controller, those parameters that are defined may be used to improve response over the corresponding controller, which lacks the model.

For example, it is believed that brainwaves represent a form of resonance, where ensembles of neurons interact in a coordinated fashion. The frequency of the wave is related to neural responsiveness to neurotransmitters, distances along neural pathways, diffusion limitations, etc. That is, the same sleep stage may be represented by slightly different frequencies in two different individuals, based on differences in the size of their brains, neuromodulators present, other anatomical, morphological and physiological differences, etc. These differences may be measured in microseconds or less, resulting in small changes in frequency. Therefore, the model component of the controller can determine the parameters of neural transmission and ensemble characteristics, vis-à-vis stimulation, and resynthesize the stimulus signal to match the correct frequency and phase of the subjects brainwave, with the optimization of the waveform adaptively determined. This may not be as simple as speeding up or slowing down playback of the signal, as different elements of the various brainwaves representing neural correlates of a sleep stage may have different relative differences between subjects.

Of course, in some cases, one or more components of the stimulation of the target subject (recipient) may be represented as abstract or semantically defined signals, and, more generally, the processing of the signals to define the stimulation will involve high-level modulation or transformation between the source signal received from the first subject (donor) or plurality of donors, to define the target signal for stimulation of the second subject (recipient).

Preferably, each component represents a subset of the neural correlates reflecting brain activity that has a high autocorrelation in space and time, or in a hybrid representation such as wavelet. These may be separated by optimal filtering (e.g., spatial PCA), once the characteristics of the signal are known, and bearing in mind that the signal is accompanied by a modulation pattern and that the two components themselves may have some weak coupling and interaction.

For example, if the first subject (donor) is listening to music, there will be significant components of the neural correlates that are synchronized with the particular music. On the other hand, the music per se may not be part of the desired stimulation of the target subject (recipient). Further, the target subject (recipient) may be in a different acoustic environment, and it may be appropriate to modify the residual signal dependent on the acoustic environment of the recipient, so that the stimulation is appropriate for achieving the desired effect, and does not represent phantoms, distractions, or irrelevant or inappropriate content. In order to perform signal processing, it is convenient to store the signals or a partially processed representation, though a complete real-time signal processing chain may be implemented. According to another embodiment, a particular stage of the sleep state of at least one first subject (donor) is identified, and the neural correlates of brain activity are captured, and the second subject (recipient) is subject to stimulation based on the captured neural correlates and the identified sleep stage. The sleep stage is typically represented as a semantic variable within a limited classification space. The sleep stage identification need not be through analysis of the neural correlates signal and may be a volitional self-identification by the first subject, e.g., based on other body signals or by an observer, or a manual classification by third parties using, for example, observation, fMRI or psychological assessment. The identified sleep stage is useful, for example, because it represents a target toward (or, in some cases, against) which the second subject (recipient) can be steered.

The stimulation may be one or more stimulus applied to the second subject (trainee or recipient), which may be a sensory stimulation (e.g., visual, auditory, or tactile), mechanical stimulation, ultrasonic stimulation, etc., and controlled with respect to waveform, frequency, phase, intensity/amplitude, duration, or controlled via feedback, self-reported effect by the second subject, manual classification by third parties, automated analysis of brain activity, behavior, physiological parameters, etc. of the second subject (recipient).

Typically, the goal of the process is to improve sleep in a recipient by transplanting the desired sleep stages, or a sequence of stages, of at least one first subject (donor) to the second subject (recipient) by inducing in the second subject (recipient) neural correlates of the sleep stage (or a sequence of stages) of at least one first subject (donor) corresponding to the sleep stage of the first subject, through the use of stimulation parameters comprising a waveform over a period of time derived from the neural correlates of the sleep stage of the first subject.

Typically, the first and the second subjects are spatially remote from each other and may be temporally remote as well. In some cases, the first and second subject are the same subject (human or animal), temporally displaced. In other cases, the first and the second subject are spatially proximate to each other. These different embodiments differ principally in the transfer of the signal from at least one first subject (donor) to the second subject (recipient). However, when the first and the second subjects share a common environment, the signal processing of the neural correlates and, especially of real-time feedback of neural correlates from the second subject, may involve interactive algorithms with the neural correlates of the first subject.

According to another embodiment, the first and second subjects are each subject to stimulation. In one particularly interesting embodiment, the first subject and the second subject communicate with each other in real-time, with the first subject receiving stimulation based on the second subject, and the second subject receiving feedback based on the first subject. This can lead to synchronization of neural correlates (e.g., neuronal oscillations, or brainwaves) and, consequently, of sleep stage between the two subjects. The neural correlates may be neuronal oscillations resulting in brainwaves that are detectable as, for example, EEG, qEEG, or MEG signals. Traditionally, these signals are found to have dominant frequencies, which may be determined by various analyses, such as spectral analysis, wavelet analysis, or principal component analysis (PCA), for example. One embodiment provides that the modulation pattern of a brainwave of at least one first subject (donor) is determined independently of the dominant frequency of the brainwave (though, typically, within the same class of brainwaves), and this modulation imposed on a brainwave corresponding to the dominant frequency of the second subject (recipient). That is, once the second subject achieves that same brainwave pattern as the first subject (which may be achieved by means other than electromagnetic, mechanical, or sensory stimulation), the modulation pattern of the first subject is imposed as a way of guiding the sleep stage of the second subject.

According to another embodiment, the second subject (recipient) is stimulated with a stimulation signal, which faithfully represents the frequency composition of a defined component of the neural correlates of at least one first subject (donor). The defined component may be determined based on principal component analysis, independent component analysis (ICI), eigenvector-based multivariable analysis, factor analysis, canonical correlation analysis (CCA), nonlinear dimensionality reduction (NLDR), or related technique.

The stimulation may be performed, for example, by using a light stimulation, a sound stimulation, a tactile stimulation, or olfactory stimulation. An auditory stimulus may be, for example, binaural beats or isochronic tones. Non-sensory stimulation may include a TES device, such as a tDCS device, a high-definition tDCS device, an osc-tDCS device, a pulse-tDCS (“electrosleep”) device, an osc-tDCS, a tACS device, a CES device, a TMS device, rTIMS device, a deep TMS device, a light source, or a sound source configured to modulate the dominant frequency on respectively the light signal or the sound signal. The stimulus may be a light signal, a sonic signal (sound), an electric signal, a magnetic field, olfactory or tactile stimulation. The current signal may be a pulse signal or an oscillating signal. The stimulus may be applied via a cranial electric stimulation (CES), a transcranial electric stimulation (TES), a deep electric stimulation, a transcranial magnetic stimulation (TMS), a deep magnetic stimulation.

The technology also provides a processor configured to process the neural correlates of sleep stage from the first subject (donor), and to produce or define a stimulation pattern for the second subject (recipient) selectively dependent on a waveform pattern of the neural correlates from the first subject. Typically, the processor performs signal analysis and calculates at least a dominant frequency of the brainwaves of the first subject, and preferably also spatial and phase patterns within the brain of the first subject. The processor may also perform a PEA, a spatial PEA, an independent component analysis (ICA), eigenvalue decomposition, eigenvector-based multivariate analyses, factor analysis, an autoencoder neural network with a linear hidden layer, linear discriminant analysis, network component analysis, nonlinear dimensionality reduction (NLDR), or another statistical method of data analysis.

A signal is presented to a second apparatus, configured to stimulate the second subject (recipient), which may be an open loop stimulation dependent on a non-feedback-controlled algorithm, or a closed loop feedback dependent algorithm. The second apparatus produces a stimulation intended to induce in the second subject (recipient) the desired sleep stage, e.g., representing the same sleep stage as was present in the first subject (donor).

A typically process performed on the neural correlates is filtering to remove noise. In some embodiments, noise filters may be provided, for example, at 50 Hz, 60 Hz, 100 Hz, 120 Hz, and additional overtones (e.g., tertiary and higher harmonics). The stimulator associated with the second subject (recipient) would typically perform decoding, decompression, decryption, inverse transformation, modulation, etc.

Alternately, an authentic wave or hash thereof may be authenticated via a blockchain, and thus authenticatable by an immutable record. In some cases, it is possible to use the stored encrypted signal in its encrypted form, without decryption. For example, with an asymmetric encryption scheme, which supports distance determination.

Due to different brain sizes, and other anatomical, morphological, and/or physiological differences, dominant frequencies associated with the same sleep stage may be different in different subjects. Consequently, it may not be optimal to forcefully impose on the recipient the frequency of the donor that may or may not precisely correspond to the recipient's frequency associated with the same sleep stage. Accordingly, in some embodiments, the donors frequency may be used to start the process of inducing the desired sleep stage in a recipient. As some point, when the recipient is closed to achieving the desired sleep state, the stimulation is either stopped or replaced with neurofeedback allowing the brain of the recipient to find its optimal frequency associated with the desired sleep stage.

In one embodiment, the feedback signal from the second subject may be correspondingly encoded as per the source signal, and the error between the two minimized. According to one embodiment, the processor may perform a noise reduction distinct from frequency-band filtering. According to one embodiment, the neural correlates are transformed into a sparse matrix, and in the transform domain, components having a high probability of representing noise are masked, while components having a high probability of representing signal are preserved. That is, in some cases, the components that represent modulation that are important may not be known a priori. However, dependent on their effect in inducing the desired response in the second subject (recipient), the “important” components may be identified, and the remainder filtered or suppressed. The transformed signal may then be inverse-transformed and used as a basis for a stimulation signal.

According to another embodiment, a method of sleep stage modification, e.g., brain entrainment, is provided, comprising: ascertaining a sleep stage in a plurality of first subjects (donors); acquiring brainwaves of the plurality of first subjects (donors), e.g., using one of EEG and MEG, to create a dataset containing brainwaves corresponding to different sleep stages. The database may be encoded with a classification of sleep stages, activities, environment, or stimulus patterns, applied to the plurality of first subjects, and the database may include acquired brainwaves across a large number of sleep stages, activities, environment, or stimulus patterns, for example. In many cases, the database records will reflect a characteristic or dominant frequency of the respective brainwaves.

Sleep stages, activities, environment, or stimulus patterns, for example, and a stimulation pattern for a second subject (recipient) defined based on the database records of one or more subjects (donors).

The record(s) thus retrieved are used to define a stimulation pattern for the second subject (recipient). As a relatively trivial example, a female recipient could be stimulated principally based on records from female donors. Similarly, a child recipient of a certain age could be stimulated principally based on the records from children donors of a similar age. Likewise, various demographic, personality, and/or physiological parameters may be matched to ensure a high degree of correspondence to between the source and target subjects. In the target subject, a guided or genetic algorithm may be employed to select modification parameters from the various components of the signal, which best achieve the desired target state based on feedback from the target subject.

Of course, a more nuanced approach is to process the entirety of the database and stimulate the second subject based on a global brainwave-stimulus model, though this is not required, and also, the underlying basis for the model may prove unreliable or inaccurate. It may be preferred to derive a stimulus waveform from only a single first subject (donor), in order to preserve micro-modulation aspects of the signal, which, as discussed above, have not been fully characterized. However, the selection of the donor(s) need not be static and can change frequently. The selection of donor records may be based on population statistics of other users of the records, i.e., whether or not the record had the expected effect, filtering donors whose response pattern correlates highest with a given recipient, etc. The selection of donor records may also be based on feedback patterns from the recipient.

The process of stimulation typically seeks to target the desired sleep stage in the recipient, which is automatically or semi-automatically determined or manually entered. In one embodiment, the records are used to define a modulation waveform of a synthesized carrier or set of carriers, and the process may include a frequency domain multiplexed multi-subcarrier signal (which is not necessarily orthogonal). A plurality of stimuli may be applied concurrently, through the different subchannels and/or though different stimulator electrodes, electric current stimulators, magnetic field generators, mechanical stimulators, sensory stimulators, etc. The stimulus may be applied to achieve brain entrainment (i.e., synchronization) of the second subject (recipient) with one or more first subjects (donors). If the plurality of donors are mutually entrained, then each will have a corresponding brainwave pattern dependent based on brainwave entrainment. This link between donors may help determine compatibility between a respective donor and the recipient. For example, characteristic patterns in the entrained brainwaves may be determined, even for different target sleep stages, and the characteristic patterns may be correlated to find relatively close matches and to exclude relatively poor matches.

This technology may also provide a basis for a social network, dating site, employment, mission (e.g., space or military), or vocational testing, or other interpersonal environments, wherein people may be matched with each other based on entrainment characteristics. For example, people who efficiently entrain with each other may have better compatibility and, therefore, a better marriage, work, or social relationships than those who do not. The entrainment effect need not be limited to sleep stages and may arise across any context.

As discussed above, the plurality of first subjects (donors) may have their respective brainwave patterns stored in separate database records. However, they may also be combined into a more global model. One such model is a neural network or a deep neural network. Typically, such a network would have recurrent features. Data from a plurality of first subjects (donors) is used to train the neural network, which is then accessed by inputting the target stage and/or feedback information, and which outputs a stimulation pattern or parameters for controlling a stimulator(s). When multiple first subjects (donors) form the basis for the stimulation pattern, it is preferred that the neural network output parameters of the stimulation, derived from and comprising features of the brainwave patterns or other neural correlates of sleep stage from the plurality of first subject (donors), which are then used to control a stimulator which, for example, generates its own carrier wave(s) which are then modulated based on the output of the neural network. A trained neural network need not periodically retrieve records and, therefore, may operate in a more time-continuous manner, rather than the more segmented scheme of record-based control.

In any of the feedback dependent methods, the brainwave patterns or other neural correlates of sleep stages may be processed by a neural network, to produce an output that guides or controls the stimulation. The stimulation, is, for example, at least one of a light signal, a sound signal, an electric signal, a magnetic field, an olfactory signal, a chemical signal, and vibration or mechanical stimulus. The process may employ a relational database of sleep stages and brainwave patterns, e.g., frequencies/neural correlate waveform patterns associated with the respective sleep stages. The relational database may comprise a first table, the first table further comprising a plurality of data records of brainwave patterns, and a second table, the second table comprising a plurality of sleep stages, each of the sleep stages being linked to at least one brainwave pattern. Data related to sleep stages and brainwave patterns associated with the sleep stages are stored in the relational database and maintained. The relational database is accessed by receiving queries for selected (existing or desired) sleep stages, and data records are returned representing the associated brainwave pattern. The brainwave pattern retrieved from the relational database may then be used for modulating a stimulator seeking to produce an effect selectively dependent on the desired sleep stage.

A further aspect of the technology provides a computer apparatus for creating and maintaining a relational database of sleep stages and frequencies associated with the sleep stage. The computer apparatus may comprise a non-volatile memory for storing a relational database of sleep stages and neural correlates of brain activity associated with the sleep stages, the database comprising a first table comprising a plurality of data records of neural correlates of brain activity associated with the sleep stages, and a second table comprising a plurality of sleep stages, each of the sleep stages being linked to one or more records in the first table; a processor coupled with the non-volatile memory, and is configured to process relational database queries, which are then used for searching the database; RAM coupled with the processor and the non-volatile memory for temporary holding database queries and data records retrieved from the relational database; and an IO interface configured to receive database queries and deliver data records retrieved from the relational database. A structured query language (SQL) or alternate to SQL (e.g., noSQL) database may also be used to store and retrieve records. A relational database described above, maintained and operated by a general-purpose computer, improves the operations of the general-purpose computer by making searches of specific sleep stages and brainwaves associated therewith more efficient thereby, laterals, reducing the demand on computing power.

A further aspect of the technology provides a method of brain entrainment comprising: ascertaining a sleep stage in at least one first subject (donor), recording brainwaves of said at least one first subject (donor) using at least one channel of EEG and/or MEG; storing the recorded brainwaves in a physical memory device, retrieving the brainwaves from the memory device, applying a stimulus signal comprising a brainwave pattern derived from at least one-channel of the EEG and/or MEG to a second subject (recipient) via sensory stimulation, whereby the sleep stage desired by the second subject (recipient) is achieved. The stimulation may be of the same dimension (number of channels) as the EEG or MEG, or a different number of channels, typically reduced. For example, the EEG or MEG may comprise 64, 128, or 256 channels.

One of the advantages of transforming the data is the ability to select a transform that separates the information of interest represented in the raw data, from noise or other information. Some transforms preserve the spatial and state transition history and may be used for a more global analysis. Another advantage of a transform is that it can present the information of interest in a form where relatively simple linear or statistical functions of a low order may be applied. In some cases, it is desired to perform an inverse transform on the data. For example, if the raw data includes noise, such as 50 or 60 Hz interference, a frequency transform may be performed, followed by a narrow band filtering of the interference and its higher order intermodulation products. An inverse transform may be performed to return the data to its time-domain representation for further processing. (In the case of simple filtering, a finite impulse response (FIR) or infinite impulse response (IIR) filter could be employed). In other cases, the analysis is continued in the transformed domain.

Transforms may be part of an efficient algorithm to compress data for storage or analysis, by making the representation of the information of interest consume fewer bits of information (if in digital form) and/or allow it to be communicated using lower bandwidth. Typically, compression algorithms will not be lossless, and as a result, the compression is irreversible with respect to truncated information.

Typically, the transformation(s) and filtering of the signal are conducted using traditional computer logic, according to defined algorithms. The intermediate stages may be stored and analyzed. However, in some cases, neural networks or deep neural networks may be used, convolutional neural network architectures, or even analog signal processing. According to one set of embodiments, the transforms (if any) and analysis are implemented in a parallel processing environment. Such as using a SIMD processor such as a GPU (or GPGPU). Algorithms implemented in such systems are characterized by an avoidance of data-dependent branch instructions, with many threads concurrently executing the same instructions.

EEG signals are analyzed to determine the location (e.g., voxel or brain region) from which an electrical activity pattern is emitted, and the wave pattern characterized. The spatial processing of the EEG signals will typically precede the content analysis, since noise and artifacts may be useful for spatial resolution. Further, the signal from one brain region will typically be noise or interference in the signal analysis from another brain region; so the spatial analysis may represent part of the comprehension analysis. The spatial analysis is typically in the form of a geometrically and/or anatomically-constrained statistical model, employing all of the raw inputs in parallel. For example, where the input data is transcutaneous electroencephalogram information, from 32 EEG electrodes, the 32 input channels, sampled at e.g., 500 sps, 1 ksps or 2 ksps, are processed in a four or higher dimensional matrix, to permit mapping of locations and communication of impulses overtime, space and state.

The matrix processing may be performed in a standard computing environment, e.g., an i9-9900K, i9-9980HK, processor, under the Windows 10 operating system, executing Matlab (Mathworks, Woburn Mass.) software platform. Alternately, the matrix processing may be performed in a computer duster or grid or cloud computing environment. The processing may also employ parallel processing, in either a distributed and loosely coupled environment, or asynchronous environment. One preferred embodiment employs a single instruction, multiple data processors, such as a graphics processing unit such as the nVidia CUDA environment or AMD Firepro high-performance computing environment.

Artificial intelligence (AI) and machine learning methods, such as artificial neural networks, deep neural networks, etc., may be implemented to extract the signals of interest. Neural networks act as an optimized statistical classifier and may have arbitrary complexity. A so-called deep neural network having multiple hidden layers may be employed. The processing is typically dependent on labeled training data, such as EEG data, or various processed, transformed, or classified representations of the EEG data. The label represents the emotion, mood, context, or state of the subject during acquisition. In order to handle the continuous stream of data represented by the EEG, a recurrent neural network architecture may be implemented. Depending preprocessing before the neural network, formal implementations of recurrence may be avoided. A four or more dimensional data matrix may be derived from the traditional spatial-temporal processing of the EEG and fed to a neural network. Since the time parameter is represented in the input data, a neural network temporal memory is not required, though this architecture may require a larger number of inputs. Principal component analysis (PCA, en.wikipedia.org/wiki/Principal_component_analysis), spatial PCA (arxiv.org/pdf/1501.03221v3.pdf, adegenetr-forgar-projectorg/files/tutorial-spca.pdf, mmi.ncbi.nlm.nih.gov/pubmed/1510870); and clustering analysis may also be employed (en.wikipedia.org/wiki/Cluster analysis, see U.S. Pat. Nos. 9,336,302, 9,607,023 and cited references).

In general, a neural network of this type of implementation will, in operation, be able to receive unlabeled EEG data, and produce the output signals representative of the predicted or estimated task, performance, context, or state of the subject during the acquisition of the unclassified EEG. Of course, statistical classifiers may be used rather than neural networks.

The analyzed EEG, either by conventional processing, neural network processing, or both, serves two purposes. First, it permits one to deduce which areas of the brain are subject to which kinds of electrical activity under which conditions. Second, it permits feedback during the training of a trainee (assuming proper spatial and anatomical correlates between the trainer and trainee), to help the system achieve the desired state, or as may be appropriate, the desired series of states and/or state transitions. According to one aspect of the technology, the applied stimulation is dependent on a measured starting state or status (which may represent a complex context and history-dependent matrix of parameters), and therefore the target represents a desired complex vector change. Therefore, this aspect of the technology seeks to understand a complex time-space-brain activity associated with an activity or task in a trainer, and to seek a corresponding complex time-space-brain activity associated with the same activity or task in a trainee, such that the complex time-space-brain activity state in the trainor is distinct from the corresponding state sought to be achieved in the trainee. This permits the transfer of training paradigms from qualitatively different persons, in different contexts, and, to some extent, to achieve a different result.

The conditions of data acquisition from the trainer will include both task data, and sensory-stimulation data. That is, a preferred application of the system is to acquire EEG data from a trainer or skilled individual, which will then be used to transfer learning, or more likely, learning readiness states, to a naïve trainee. The goal for the trainee is to produce a set of stimulation parameters that will achieve, in the trainee, the corresponding neural activity resulting in the EEG state of the trainer at the time of or preceding the learning of a skill or a task, or performance of the task.

It is noted that EEG is not the only neural or brain activity or state data that may be acquired, and, of course, any and all such data may be included within the scope of the technology, and therefore EEG is a representative example only of the types of data that may be used. Other types include fMRI, magnetoencephalogram, motor neuron activity, PET, etc.

While mapping the stimulus-response patterns distinct from the task is not required in the trainer, it is advantageous to do so, because the trainer may be available for an extended period, the stimulus of the trainee may influence the neural activity patterns, and it is likely that the trainer will have correlated stimulus-response neural activity patterns with the trainee(s). It should be noted that the foregoing has suggested that the trainer is a single individual, while in practice, the trainer may be a population of trainers or skilled individuals. The analysis and processing of brain activity data may, therefore, be adaptive, both for each respective individual and for the population as a whole.

For example, the system may determine that not all human subjects have common stimulus-response brain activity correlates, and therefore that the population needs to be segregated and clustered. If the differences may be normalized, then a normalization matrix or other correction may be employed. On the other hand, if the differences do not permit feasible normalization, the population(s) may be segmented, with different trainers for the different segments. For example, in some tasks, male brains have different activity patterns and capabilities than female brains. This, coupled with anatomical differences between the sexes, implies that the system may provide gender-specific implementations. Similarly, age differences may provide a rational and scientific basis for segmentation of the population. However, depending on the size of the information base and matrices required, and some other factors, each system may be provided with substantially all parameters required for the whole population, with a user-specific implementation based on a user profile or initial setup, calibration, and system training session.

According to one aspect of the present invention, a source subject is instrumented with sensors to determine localized brain activity during experiencing an event. The objective is to identify regions of the brain involved in processing this response.

The sensors will typically seek to determine neuron firing patterns and brain region excitation patterns, which can be detected by implanted electrodes, transcutaneous electroencephalograms, magnetoencephalograms, fMRI, and other technologies. Where appropriate, transcutaneous EEG is preferred, since this is non-invasive and relatively simple.

The source is observed with the sensors in a quiet state, a state in which he or she is experiencing an event, and various control states in which the source is at rest or engaged in different activities resulting in different states. The data may be obtained for a sufficiently long period of time and over repeated trials to determine the effect of duration. The data may also be a population statistical result, and need not be derived from only a single individual at a single time.

The sensor data is then processed using a 4D (or higher) model to determine the characteristic location-dependent pattern of brain activity over time associated with the state of interest. Where the data is derived from a population with various degrees of arousal, the model maintains this arousal state variable dimension.

A recipient is then prepared for receipt of the mental state. The mental state of the recipient may be assessed. This can include responses to a questionnaire, self-assessment, or other psychological assessment methods. Further, the transcutaneous EEG (or other brain activity data) of the recipient may be obtained, to determine the starting state for the recipient, as well as an activity during experiencing the desired mental state.

In addition, a set of stimuli, such as visual patterns, acoustic patterns, vestibular, smell, taste, touch (light touch, deep touch, proprioception, stretch, hot, cold, pain, pleasure, electric stimulation, acupuncture, etc.), vagus nerve (e.g., parasympathetic), are imposed on the subject, optionally over a range of baseline brain states, to acquire data defining the effect of individual and various combinations of sensory stimulation on the brain state of the recipient. Population data may also be used for this aspect.

The data from the source or population of sources (see above) may then be processed in conjunction with the recipient or population of recipient data, to extract information defining the optimal sensory stimulation over time of the recipient to achieve the desired brain state resulting in the desired mental state.

In general, for populations of sources and recipients, the data processing task is immense. However, the statistical analysis will generally be of a form that permits parallelization of mathematical transforms for processing the data, which can be efficiently implemented using various parallel processors, a common form of which is a SIMD (single instruction, multiple data) processor, found in typical graphics processors (GPUs). Because of the cost-efficiency of GPUs, it is referred to implement the analysis using efficient parallelizable algorithms, even if the computational complexity is nominally greater than a CISC-type processor implementation.

During stimulation of the recipient, the EEG pattern may be monitored to determine if the desired state is achieved through the sensory Stimulation. A closed loop feedback control system may be implemented to modify the stimulation seeking to achieve the target. An evolving genetic algorithm may be used to develop a user model, which relates the mental state, arousal, and valence, sensory stimulation, and brain activity patterns, both to optimize the current session of stimulation and learning, as well as to facilitate future sessions, where the mental states of the recipient have further enhanced, and to permit use of the system for a range of mental states.

The technology may be embodied in apparatuses for acquiring the brain activity information from the source, processing the brain activity information to reveal a target brain activity state and a set of stimuli, which seek to achieve that state in a recipient, and generating stimuli for the recipient to achieve and maintain the target brain activity state over a period of time and potential state transitions. The generated stimuli may be feedback controlled. A general-purpose computer may be used for the processing of the information, a microprocessor, an FPGA, an ASIC, a system-on-a-chip, or a specialized system, which employs a customized configuration to efficiently achieve the information transformations required. Typically, the source and recipient act asynchronously, with the brain activity of the source recorded and later processed. However, real-time processing and brain activity transfer are also possible. In the case of a general purpose programmable processor implementation or portions of the technology, computer instructions may be stored on a nontransient computer readable medium. Typically, the system will have special-purpose components, such as a sensory stimulator, or a modified audio and/or display system, and therefore the system will not be a general purpose system. Further, even in a general purpose system, the operation per se is enhanced according to the present technology.

Mental states may be induced in a subject non-invasively via light, sound, or other means capable of conveying frequency patterns.

The transmission of the brainwaves can be accomplished through direct electrical contact with the electrodes implanted in the brain or remotely employing light, sound, electromagnetic waves, and other non-invasive techniques. Light, sound, or electromagnetic fields may be used to remotely convey the temporal pattern of prerecorded brainwaves to a subject by modulating the encoded temporal frequency on the light, sound or electromagnetic filed signal to which the subject is exposed.

Every activity, mental or motor, and emotion is associated with unique brainwaves having specific spatial and temporal patterns, i.e., a characteristic frequency or a characteristic distribution of frequencies over time and space. Such waves can be read and recorded by several known techniques, including electroencephalography (EEG), magnetoencephalography (MEG), exact low-resolution brain electromagnetic tomography (eLORETA), sensory evoked potentials (SEP), fMRI, functional near-infrared spectroscopy (fNIRS), etc. The cerebral cortex is composed of neurons that are interconnected in networks. Cortical neurons constantly send and receive nerve impulses-electrical activity-even during sleep. The electrical or magnetic activity measured by an EEG or MEG (or another device) device reflects the intrinsic activity of neurons in the cerebral cortex and the information sent to it by subcortical structures and the sense receptors.

It has been observed that “playing back the brainwaves” to another animal or person by providing decoded temporal pattern through transcranial direct current stimulation (tDCS), transcranial alternating current stimulation (tACS), high definition transcranial alternating current stimulation (HD-tDCS), transcranial magnetic stimulation (TMS), or through electrodes implanted in the brain allows the recipient to achieve the mental state at hand or to increase the speed of achievement. For example, if the brainwaves of a mouse navigated a familiar maze are decoded (by EEG or via implanted electrodes), playing this temporal pattern to another mouse unfamiliar with this maze will allow it to learn to navigate this maze faster.

Similarly, recording brainwaves associated with a specific response of one subject and late “playing back” this response to another subject will induce a similar response in the second subject. More generally, when one animal assumes a mental state, parts of the brain will have characteristic activity patterns. Further, by “artificially” inducing the same pattern in another animal, the other animal will have the same mental state or more easily be induced into that state. The pattern of interest may reside deep in the brain, and thus be overwhelmed in an EEG signal by cortical potentials and patterns. However, techniques other than surface electrode EEG may be used to determine and spatially discriminate deep brain activity, e.g., from the limbic system. For example, various types of magnetic sensors may sense deep brain activity. See, e.g., U.S. Pat. Nos. 9,618,591; 9,261,573; 8,618,799; and 8,593,141.

In some cases, EEGs dominated by cortical excitation patterns may be employed to sense the mental state, since the cortical patterns may correlate with lower-level brain activity. Note that the determination of a state representation of a mental state need not be performed each time the system is used; rather, once the brain spatial and temporal activity patterns and synchronization states associated with a particular mental states are determined, those patterns may be used for multiple targets and overtime.

Similarly, while the goal is, for example, to trigger the target to assume the same brain activity patterns are the exemplar, this can be achieved in various ways, and these methods of inducing the desired patterns need not be invasive. Further, user feedback, especially in the case of a human transferee, may be used to tune the process. Finally, using the various senses, especially sight, sound, vestibular, touch, proprioception, taste, smell, vagus afferent, other cranial nerve afferent, etc. can be used to trigger high-level mental activity, that in a particular subject achieves the desired mental state, emotion or mood.

Thus, in an experimental subject, which may include laboratory scale and/or invasive monitoring, a set of brain electrical activity patterns that correspond to particular emotions or mental states is determined. Preferably, these are also correlated with surface EEG findings. For the transferee, a stimulation system is provided that is non-hazardous and non-invasive. For example, audiovisual stimulation may be exclusively used. A set of EEG electrodes is provided to measure brain activity, and an adaptive or genetic algorithm scheme is provided to optimize the audiovisual presentation, seeking to induce in the transferee the target pattern found in the experimental subject. After the stimulation patterns, which may be path dependent, are determined, it is likely that these patterns will be persistent, though over longer time periods, there may be some desensitization to the stimulation pattern(s). In some cases, audiovisual stimulation is insufficient, and TMS, PEMF, or other electromagnetic stimulation (superthreshold, or preferably subthreshold) is employed to assist in achieving the desired state and maintaining it for the desired period.

Employing light, sound or electromagnetic field to remotely convey the temporal pattern of brainwaves (which may be prerecorded) to a subject by modulating the encoded temporal frequency on the light, sound or electromagnetic filed signal to which the subject is exposed.

When a group of neurons fires simultaneously, the activity appears as a brainwave. Different brainwave-frequencies are linked to different mental states in the brain.

The desired mental state may be induced in a target individual (e.g., human, animal), by providing selective stimulation according to a temporal pattern, wherein the temporal pattern is correlated with an EEG pattern of the target when in the desired mental state, or represents a transition which represents an intermediate toward achieving the desired mental state. The temporal pattern may be targeted to a discrete spatial region within the brain, either by a physical arrangement of a stimulator, or natural neural pathways through which the stimulation (or its result) passes.

The EEG pattern may be derived from another individual or individuals, the same individual at a different time, or an in vivo animal model of the desired mental state. The method may, therefore, replicate a mental state of a first subject in a second subject. The mental state typically is not a state of consciousness or an idea, but rather a subconscious (in a technical sense) state, representing an emotion, readiness, receptivity, or another state, often independent of particular thoughts or ideas. In essence, a mental state of the first subject (a “trainer” or “donor” who is in a desired mental state) is captured by recording neural correlates of the mental state, e.g., as expressed by brain activity patterns, such as EEG or MEG signals. The neural correlates of the first subject, either as direct or recorded representations, may then be used to control a stimulation of the second subject (a “trainee” or “recipient”), seeking to induce the same brain activity patterns in the second subject (recipient/trainee) as were present in the first subject (donor/trainer) to assist the second subject (recipient/trainee) to attain the desired mental state that had been attained by the donor/trainer. In an alternative embodiment, the signals from the first subject (donor/trainer) being in the first mental state are employed to prevent the second subject (recipient/trainee) from achieving a second mental state, wherein the second mental state is an undesirable one.

The source brain wave pattern may be acquired through multichannel EEG or MEG, from a human in the desired brain state. A computational model of the brain state is difficult to create. However, such a model is not required according to the present technology. Rather, the signals may be processed by a statistical process (e.g., PCA or a related technology), or a statistically trained process (e.g., a neural network). The processed signals preferably retain information regarding signal source special location, frequency, and phase. In stimulating the recipient's brain, the source may be modified to account for brain size differences, electrode locations, etc. Therefore, the preserved characteristics are normalized spatial characteristics, frequency, phase, and modulation patterns.

The normalization may be based on feedback from the target subject, for example, based on a comparison of a present state of the target subject and a corresponding state of the source subject, or another comparison of known states between the target and source. Typically, the excitation electrodes in the target subject do not correspond to the feedback electrodes or the electrodes on the source subject. Therefore, an additional type of normalization is required, which may also be based on a statistical or statistically trained algorithm.

According to one embodiment, the stimulation of the second subject is associated with a feedback process, to verify that the second subject has appropriately responded to the stimulation, e.g., has a predefined similarity to the mental state as the first subject, has a mental state with a predefined difference from the first subject, or has a desired change from a baseline mental state. Advantageously, the stimulation may be adaptive to the feedback. In some cases, the feedback may be functional, i.e., not based on brain activity per se, or neural correlates of mental state, but rather physical, psychological, or behavioral effects that may be reported or observed.

The feedback typically is provided to a computational model-based controller for the stimulator, which alters stimulation parameters to optimize the stimulation in dependence on a brain and brain state model applicable to the target.

For example, it is believed that brainwaves represent a form of resonance, where ensembles of neurons interact in a coordinated fashion as a set of coupled or interacting oscillators. The frequency of the wave is related to neural responsivity to neurotransmitters, distances along neural pathways, diffusion limitations, etc., and perhaps pacemaker neurons or neural pathways. That is, the same mental state may be represented by different frequencies in two different individuals, based on differences in the size of their brains, neuromodulators present, physiological differences, etc. These differences may be measured in microseconds or less, resulting in fractional changes in frequency. However, if the stimulus is different from the natural or resonant frequency of the target process, the result may be different from that expected. Therefore, the model-based controller can determine the parameters of neural transmission and ensemble characteristics, vis-à-vis stimulation, and resynthesize the stimulus wave to match the correct waveform, with the optimization of the waveform adaptively determined. This may not be as simple as speeding up or slowing down playback of the signal, as different elements of the various waveforms representing neural correlates of mental state may have different relative differences between subjects. Therefore, according to one set of embodiments, the stimulator autocalibrates for the target, based on correspondence (error) of a measured response to the stimulation and the desired mental state sought by the stimulation. In cases where the results are chaotic or unpredictable based on existing data, a genetic algorithm may be employed to explore the range of stimulation parameters, and determine the response of the target.

In some cases, the target has an abnormal or unexpected response to stimulation based on a model maintained within the system. In this case, when the deviance from the expected response is identified, the system may seek to a new model, such as from a model repository that may be online, such as through the Internet. If the models are predictable, a translation may be provided between an applicable model of a source or trainer, and the applicable model of the target, to account for differences. In some cases, the desired mental state is relatively universal, such as sleep and awake. In this case, the brain response model may be a statistical model, rather than a neural network or deep neural network type implementation.

Thus, in one embodiment, a hybrid approach is provided, with the use of donor-derived brainwaves, on the one hand, which may be extracted from the brain activity readings (e.g., EEG or MEG) of the first at least one subject (donor), preferably processed by principal component analysis, or spatial principal component analysis, autocorrelation, or other statistical processing technique (clustering, PCA, etc) or statistically trained technique (backpropagation of errors, etc.) that separates components of brain activity, which can then be modified or modulated based on high-level parameters, e.g., abstractions. See, ml4a.github.io/m14a/how_neural_networksaretrained/. Thus, the stimulator may be programmed to induce a series of brain states defined by name (e.g., sleep stage 1, sleep stage 2, etc.) or as a sequence of “abstract” semantic labels, icons, or other representations, each corresponding to a technical brain state or sequence of sub-states. The sequence may be automatically defined, based on biology and the system training, and thus relieve the programmer of low-level tasks. However, in a general case, the present technology maintains the use of components or subcomponents of the donors brain activity readings, e.g., EEG or MEG, and does not seek to characterize or abstract them to a semantic level.

According to the present technology, a neural network system or statistical classifier may be employed to characterize the brain wave activity and/or other data from a subject. In addition to the classification or abstraction, a reliability parameter is presented, which predicts the accuracy of the output. Where the accuracy is high, a model-based stimulator may be provided to select and/or parameterize the model and generate a stimulus for a target subject. Where the accuracy is low, a filtered representation of the signal may be used to control the stimulator, bypassing the model(s). The advantage of this hybrid scheme is that when the model-based stimulator is employed, many different parameters may be explicitly controlled independently of the source subject. On the other hand, where the data processing fails to yield a highly useful prediction of the correct model-based stimulator parameters, the model itself may be avoided, in favor of a direct stimulation type system.

Of course, in some cases, one or more components of the stimulation of the target subject may be represented as abstract or semantically defined signals, and more generally the processing of the signals to define the stimulation will involve high-level modulation or transformation between the source signal received from the first subject, to define the target signal for stimulation of the second subject.

Preferably, each component represents a subset of the neural correlates reflecting brain activity that has a high spatial autocorrelation in space and time, or in a hybrid representation such as wavelet. For example, one signal may represent a modulated 10.2 Hz signal, while another signal represents a superposed modulated 15.7 Hz signal, with respectively different spatial origins. These may be separated by optimal filtering, once the spatial and temporal characteristics of the signal are known, and bearing in mind that the signal is accompanied by a modulation pattern, and that the two components themselves may have some weak coupling and interaction.

In some cases, the base frequency, modulation, coupling, noise, phase jitter, or another characteristic of the signal may be substituted. For example, if the first subject is listening to music, there will be significant components of the neural correlates that are synchronized with the particular music. On the other hand, the music per se may not be part of the desired stimulation of the target subject. Therefore, through signal analysis and decomposition, the components of the signal from the first subject, which have a high temporal correlation with the music, may be extracted or suppressed from the resulting signal. Further, the target subject may be in a different acoustic environment, and it may be appropriate to modify the residual signal dependent on the acoustic environment of the target subject, so that the stimulation is appropriate for achieving the desired effect, and does not represent phantoms, distractions, or irrelevant or inappropriate content. In order to perform processing, it is convenient to store the signals or a partially processed representation, though a complete real-time signal processing chain may be implemented. Such a real-time signal processing chain is generally characterized in that the average size of a buffer remains constant, i.e., the lag between output and input is relatively constant, bearing in mind that there may be periodicity to the processing.

The mental state of the first subject may be identified, and the neural correlates of brain activity captured. The second subject is subject to stimulation based on the captured neural correlates and the identified mental state. The mental state may be represented as a semantic variable, within a limited classification space. The mental state identification need not be through analysis of the neural correlates signal and may be a volitional self-identification by the first subject, a manual classification by third parties, or an automated determination. The identified mental state is useful, for example, because it represents a target toward (or against) which the second subject can be steered.

The stimulation may be one or more inputs to the second subject, which may be a sensory stimulation, mechanical stimulation, ultrasonic stimulation, etc., and controlled with respect to waveform, intensity/amplitude, duration, feedback, self-reported effect by the second subject, manual classification by third parties, automated analysis of brain activity, behavior, physiological parameters, etc. of the second subject.

The process may be used to induce in the target subject neural correlates of the desired mental state, which are derived from a different time for the same person, or a different person at the same or a different time. For example, one seeks to induce the neural correlates of the first subject in a desired mental state in a second subject, through the use of stimulation parameters comprising a waveform over a period of time derived from the neural correlates of the mental state of the first subject.

The first and second subjects may be spatially remote from each other and may be temporally remote as well. In some cases, the first and second subject are the same animal (e.g., human), temporally displaced. In other cases, the first and second subject are spatially proximate to each other. In some cases, neural correlates of a desired mental state are derived from a mammal having a simpler brain, which are then extrapolated to a human brain. (Animal brain stimulation is also possible, for example, to enhance training and performance). When the first and second subjects share a common environment, the signal processing of the neural correlates, and especially of real-time feedback of neural correlates from the second subject may involve interactive algorithms with the neural correlates of the first subject.

The first and second subjects may each be subject to stimulators. The first subject and the second subject may communicate with each other in real-time, with the first subject receiving stimulation based on the second subject, and the second subject receiving feedback based on the first subject. This can lead to synchronization of mental state between the two subjects. However, the first subject need not receive stimulation based on real-time signals from the second subject, as the stimulation may derive from a third subject or the first or second subjects at different points in time.

The neural correlates may be, for example, EEG, qEEG, or MEG signals. Traditionally, these signals are found to have dominant frequencies, which may be determined by various analyses. One embodiment provides that the modulation pattern of a brainwave of the first subject is determined independent of the dominant frequency of the brainwave (though typically within the same class of brainwaves), and this modulation imposed on a wave corresponding to the dominant frequency of the second subject. That is, once the second subject achieves that same brainwave pattern as the first subject (which may be achieved by means other than electromagnetic, mechanical, or sensors stimulation), the modulation pattern of the first subject is imposed as a way of guiding the mental state of the second subject.

The second subject may be stimulated with a stimulation signal, which faithfully represents the frequency composition of a defined component of the neural correlates of the first subject.

The stimulation may be performed, for example, by using a source of one of a light signal and a sound signal configured to modulate the dominant frequency on the one of a light signal and a sound signal. The stimulus may be at least one of a light signal, a sound signal, an electric signal, and a magnetic field. The stimulus may be a light stimulation or a sound stimulation. A visual stimulus may be ambient light or direct light. An auditory stimulus may be binaural beats or isochronic tones.

The technology may also provide a processor configured to process the neural correlates of mental state from the first subject, and to produce or define a stimulation pattern for the second subject selectively dependent on a waveform pattern of the neural correlates from the first subject. Typically, the processor performs signal analysis and calculates at least a dominant frequency of the brainwaves of the first subject, and preferably also spatial and phase patterns within the brain of the first subject.

A signal is presented to a second apparatus, configured to stimulate the second subject, which may be an open loop stimulation dependent on a non-feedback controlled algorithm, or a closed loop feedback dependent algorithm. In other cases, analog processing is employed in part or in whole, wherein the algorithm comprises an analog signal processing chain. The second apparatus receives information from the processor (first apparatus), typically comprising a representation of a portion of a waveform represented in the neural correlates. The second apparatus produces a stimulation intended to induce in the second subject the desired mental state, e.g., representing the same mental state as was present in the first subject.

A typical process performed on the neural correlates is filtering to remove noise. For example, notch filters may be provided at 50 Hz, 60 Hz, 100 Hz, 120 Hz, and additional overtones. Other environmental signals may also be filtered in a frequency-selective or waveform-selective (temporal) manner. Higher level filtering may also be employed, as is known in the art. The neural correlates, after noise filtering, may be encoded, compressed (lossy or losslessly), encrypted, or otherwise processed or transformed. The stimulator associated with the second subject would typically perform decoding, decompression, decryption, inverse transformation, etc.

Information security and copy protection technology, similar to that employed for audio signals, may be employed to protect the neural correlate signals from copying or content analysis before use. In some cases, it is possible to use the stored encrypted signal in its encrypted for without decryption. For example, with an asymmetric encryption scheme, which supports distance determination. See U.S. Pat. No. 7,269,277; Sahai and Waters (2005) Annual International Conference on the Theory and Applications of Cryptographic Techniques, pp. 457-473. Springer, Berlin, Heidelberg; Bringer et al. (2009) IEEE International Conference on Communications, pp. 1-6; Juels and Sudan (2006) Designs, Codes and Cryptography 2237-257; Thaker et al. (2006) IEEE International Conference on Workload Characterization, pp. 142-149; Galil et al. (1987) Conference on the Theory and Application of Cryptographic Techniques, pp. 135-155.

Because the system may act intrusively, it may be desirable to authenticate the stimulator or parameters employed by the stimulator before use. For example, the stimulator and parameters it employs may be authenticated by a distributed ledger, e.g., a blockchain. On the other hand, in a closed system, digital signatures and other hierarchical authentication schemes may be employed. Permissions to perform certain processes may be defined according to smart contracts, which automated permissions (i.e., cryptographic authorization) provided from a blockchain or distributed ledger system. Of course, centralized management may also be employed.

In practice, the feedback signal from the second subject may be correspondingly encoded as per the source signal, and the error between the two minimized. In such an algorithm, the signal sought to be authenticated is typically brought within an error tolerance of the encrypted signal before usable feedback is available. One way to accomplish this is to provide a predetermined range of acceptable authenticatable signals which are then encoded, such that authentication occurs when the putative signal matches any of the predetermined range. In the case of the neural correlates, a large set of digital hash patterns may be provided representing different signals as hash patterns. The net result is relatively weakened encryption, but the cryptographic strength may still be sufficiently high to abate the risks.

The processor may perform a noise reduction distinct from frequency-band filtering. The neural correlates may be transformed into a sparse matrix, and in the transform domain, components representing high probability noise are masked, while components representing high probability signal are preserved. The distinction may be optimized or adaptive. That is, in some cases, the components which represent modulation that are important may not be known a priori. However, dependent on their effect in inducing the desired response in the second subject, the “important” components may be identified, and the remainder filtered or suppressed. The transformed signal may then be inverse-transformed and used as a basis for a stimulation signal.

A mental state modification, e.g., brain entrainment, may be provided, which ascertains a mental state in a plurality of first subjects; acquires brainwaves of the plurality of first subjects, e.g., using one of EEG and MEG, to create a dataset containing representing brainwaves of the plurality of first subjects. The database may be encoded with a classification of mental state, activities, environment, or stimulus patterns, applied to the plurality of first subjects, and the database may include acquired brainwaves across a large number of mental states, activities, environment, or stimulus patterns, for example. In many cases, the database records will reflect a characteristic or dominant frequency of the respective brainwaves. As discussed above, the trainer or first subject is a convenient source of the stimulation parameters but is not the sole available source. The database may be accessed according to its indexing, e.g., mental states, activities, environment, or stimulus patterns, for example, and a stimulation pattern for a second subject defined based on the database records of one or more subjects.

The record(s) thus retrieved are used to define a stimulation pattern for the second subject. The selection of records, and their use, may be dependent on the second subject and/or feedback from the second subject. As a relatively trivial example, a female second subject could be stimulated principally dependent on records from female first subjects. Of course, a more nuanced approach is to process the entirety of the database and Stimulate the second subject based on a global brain wave-stimulus model, though this is not required, and also, the underlying basis for the model may prove unreliable or inaccurate. It may be preferred to derive a stimulus waveform from only a single first subject, in order to preserve micro-modulation aspects of the signal, which as discussed above have not been fully characterized. However, the selection of the first subject(s) need not be static and can change frequently. The selection of first subject records may be based on population statistics of other users of the records (i.e., collaborative filtering, i.e., whose response pattern do I correlate highest with? etc.). The selection of first subject records may also be based on feedback patterns from the second user.

The process of stimulation may seek to target a desired mental state in the second subject, which is automatically or semi-automatically determined or manually entered. That target then represents a part of the query against the database to select the desired record(s). The selection of records may be a dynamic process, and reselection of records may be feedback dependent.

The records may be used to define a modulation waveform of a synthesized carrier or set of carriers, and the process may include a frequency domain multiplexed multi-subcarrier signal (which is not necessarily orthogonal). A plurality of stimuli may be applied concurrently, through the suffered subchannels and/or through different stimulator electrodes, magnetic field generators, mechanical stimulators, sensory stimulators, etc. The stimuli for the different subchannels or modalities need not be derived from the same records.

The stimulus may be applied to achieve the desired mental state, e.g., brain entrainment of the second subject with one or more first subjects. Brain entrainment is not the only possible outcome of this process. If the plurality of first subjects is mutually entrained, then each will have a corresponding brain wave pattern dependent based on brainwave entrainment. This link between first subject may help determine compatibility between a respective first subject and the second subject. For example, characteristic patterns in the entrained brainwaves may be determined, even for different target mental states, and the characteristic patterns correlated to find relatively close matches and to exclude relatively poor matches.

This technology may also provide a basis for a social network, dating site, employment or vocational testing, or other interpersonal environments, wherein people may be matched with each other based on entrainment characteristics. For example, people who efficiently entrain with each other may have better social relationships than those who do not. Thus, rather than seeking to match people based on personality profiles, the match could be made based on the ability of each party to efficiently entrain the brainwave pattern of the other party. This enhances non-verbal communication and assists in achieving corresponding states during activities. This can be assessed by monitoring neural responses of each individual to video, and also by providing a test stimulation based on the other party's brainwave correlates of mental state, to see whether coupling is efficiently achieved. On the other hand, the technology could be used to assist in entrainment when the natural coupling is inefficient or to block coupling where the coupling is undesirable. An example of the latter is hostility; when two people are entrained in a hostile environment, emotional escalation ensures. However, if the entrainment is attenuated, undesired escalation may be impeded.

The process may employ a relational database of mental states and brainwave patterns, e.g., frequencies/neural correlate waveform patterns associated with the respective mental states. The relational database may comprise a first table, the first table further comprising a plurality of data records of brainwave patterns, and a second table, the second table comprising a plurality of mental states, each of the mental states being linked to at least one brainwave pattern. Data related to mental states and brainwave patterns associated with the mental states are stored in the relational database and maintained. The relational database is accessed by receiving queries for selected mental states, and data records are returned, representing the associated brainwave pattern. The brainwave pattern retrieved from the relational database may then be used for modulating a stimulator seeking to produce an effect selectively dependent on the mental state at issue.

A computer apparatus may be provided for creating and maintaining a relational database of mental states and frequencies associated with the mental states, the computer apparatus comprising: a non-volatile memory for storing a relational database of mental states and neural correlates of brain activity associated with the mental states, the database comprising a first table, the first table further comprising a plurality of data records of neural correlates of brain activity associated with the mental states, and a second table, the second table comprising a plurality of mental states, each of the mental states being linked to one or more records in the first table; a processor coupled with the non-volatile memory, configured to process relational database queries, which are then used for searching the database; RAM coupled with the processor and the non-volatile memory for temporary holding database queries and data records retrieved from the relational database; and an I/O interface configured to receive database queries and deliver data records retrieved from the relational database. An SQL or noSQL database may also be used to store and retrieve records.

A further aspect of the technology provides a method of brain entrainment comprising: ascertaining a mental state in a first subject; recording brainwaves of the plurality of subjects using at least one channel one of EEG and MEG; storing the recorded brainwaves in a physical memory device; retrieving the brainwaves from the memory device; applying a stimulus signal comprising a brainwave pattern derived from at least one-channel one of the EEG and MEG to a second subject via sensory stimulation, whereby the mental state desired by the second subject is achieved. The stimulation may be of the same order (number of channels) as the EEG or MEG, or a different number of channels, typically reduced. For example, the EEG or MEG may comprise 128 or 256 channels, while the sensory stimulator may have 8 or fewer channels. Transcranial stimulation of various modalities and patterns may accompany the sensory stimulation.

The present technology may be responsive to chronobiology, and in particular to the subjective sense of time. For a subject, this may be determined volitionally subjectively, but also automatically, for example by judging attention span, using, e.g., eye movements, and analyzing the persistence of brainwave patterns or other physiological parameters after a discrete stimulus. Further, time-constants of the brain, reflected by delays and phase may also be analyzed. Further, the contingent negative variation (CNV) preceding a volitional act may be used, both to determine (or measure) conscious action timing, and also the time relationships between thought and action more generally.

Typically, brainwave activity is measured with a large number of EEG electrodes, which each receive signals from a small area on the scalp, or in the case of a MEG, by several sensitive magnetic field detectors, which are responsive to local field differences. Typically, the brainwave capture is performed in a relatively high number of spatial dimensions, e.g., corresponding to the number of sensors. It is often unfeasible to process the brainwave signals to create a source model, given that the brainwaves are created by billions of neurons, connected through axons, which have long distances. Further, the neurons are generally non-linear and interconnected. However, a source model is not required.

Various types of artificial intelligence techniques may be exploited to analyze the neural correlates of a sleep stage represented in the brain activity data of both the first subject (donor) (or a plurality of donors) and the second subject (recipient). The algorithm or implementation need not be the same, though in some cases, it is useful to confirm the approach of the source processing and feedback processing so that the feedback does not achieve or seek a suboptimal target sleep stage. However, given the possible differences in conditions, resources, equipment, and purpose, there is no necessary coordination of these processes. The artificial intelligence may take the form of neural networks or deep neural networks, though rule/expert-based systems, hybrids, and more classical statistical analysis may be used. In a typical case, an artificial intelligence process will have at least one aspect, which is non-linear in its output response to an input signal, and thus at least the principle of linear superposition is violated. Such systems tend to permit discrimination, since a decision and the process of decision-making are, ultimately, non-linear. An artificially intelligent system requires a base of experience or information upon which to train. This can be supervised (external labels applied to data), unsupervised (self-discrimination of classes), or semi-supervised (a portion of the data is externally labeled).

A self-learning or genetic algorithm may be used to tune the system, including both or either the signal processing at the donor system and the recipient system. In a genetic algorithm feedback-dependent self-learning system, the responsivity of a subject, e.g., the target, to various kinds of stimuli may be determined over a stimulus space. This stimulation may be in the context of use, with a specific target sleep stage provided, or unconstrained. The stimulator may operate using a library of stimulus patterns, or seek to generate synthetic patterns or modifications of patterns. Over some time, the system will learn to map the desired sleep stage to optimal context-dependent parameters of the stimulus pattern.

The technology may be used for both the creation of a desired sleep stages in the recipient, elimination of existing sleep stages in the recipient. In the latter case, a decision of what end state is to be achieved is less constrained, and therefore, the optimization is distinct. For example, in the former case, it may be hard to achieve a particular sleep stage that is desired, requiring a set of transitions to cause the brain of the recipient to be enabled/prepared to enter the target state. In the case of a system seeking to eliminate an undesired sleep stage, the issue is principally what path to take to most efficiently leave the current state, bearing in mind the various costs, such as the comfort/discomfort of the stimulation, the time value cost, etc. Therefore, the series of states may differ in the implementation of these distinct goals, even if the endpoints are identical, i.e., the optimal algorithm to achieve state B from state A, may be different from the optimal algorithm to exist state A, and end up at state B.

The technology may be used to address sleep stages or sections of them associated with dreaming. Typically, dreaming is associated with many different brain regions. As such, the biology of dreaming is different. Often, dreams have a biochemical or hormonal component and, perhaps, a physiological component, that may be attenuated or absent from cognitive states. Dreaming had long been thought to occur largely during rapid-eye-movement (REM) sleep, but dreams have also been reported to occur during non-REM sleep. However, dreams are typically remembered, if the dreamer wakes us during the REM phase of the sleep. In addition, it has been shown that dreaming, for example, about faces was linked to increased high-frequency activity in the specific region of the brain involved in face recognition, with dreams involving spatial perception, movement and thinking similarly linked to regions of the brain that handle such tasks when awake. Therefore, while the general brainwave or other neural correlates acquisition from a donor may be similar or identical, the stimulus used on the second subject (recipient) may be distinct in modality, spatial location, intensity/waveform, other stimulation parameters, and the types and application of feedback employed.

It is known that people who have more REM sleep and more intense theta (4 Hz-7 Hz) activity during REM are better able to consolidate emotional memories. It was suggested (Blagrove) that if we attempt to hack our dreams by artificially increasing theta waves, it might lead to the incorporation of more waking experiences into our dreams. (See “Dreams act as overnight therapy” New Scientist magazine on 5 May 2018). Transplanting theta frequency brainwaves from a vivid dreamer may also help achieve the same effect. Moreover, instead of stimulating the subjects brain with a synthetic theta frequency (e.g., isotonic tones or ambient sound beats), stimulating the recipient's brain using donors brainwaves carrying secondary (and higher) harmonics, in addition to the dominant theta frequency, may induce the same category of dreams, i.e., if the donor dreamed of people, the recipient will be more likely to dream of people, albeit different people, because the donors brainwaves will stimulate the visual cortex of the recipient. This may be helpful in the treatment of MD, stress management, phobias and some psychiatric diseases.

In a medical treatment implementation, in some cases it may be appropriate to administer a drug or pharmacological agent, such as melatonin, hypnotic or soporific drug, a sedative (e.g., barbiturates, benzodiazepines, nonbenzodiazepine hypnotics, orexin antagonists, antihistamines, general anesthetics, cannabis and other herbal sedatives, methaqualone and analogues, muscle relaxants, opioids) that assists in achieving the target sleep stage, and for emotional states and/or dreams, this may include certain psychotropic drugs, such as epinephrine, norepinephrine reuptake inhibitors, serotonin reuptake inhibitors, peptide endocrine hormones, such as oxytocin, ACTH fragments, insulin, etc. Combining a drug with stimulation may reduce the required dose of the drug and the associated side effects of the drug.

It is an object to provide a method of brain entrainment to facilitate sleep in a subject using a sleep app executing on a user device, the method comprising: executing the sleep app on the user device, configured to select at least one stimulus selected from the group consisting of at least one of a light signal and a sound signal; selecting a waveform from a menu having a plurality of waveforms derived from brainwaves of at least one sleeping donor, wherein said waveform corresponds to at least one specific stage of sleep; and stimulating the subject with said at least one stimulus, wherein said at least one stimulus is modulated with the selected waveform, to thereby entrain the brain of the subject with the selected waveform to facilitate sleep in the subject. The user device may be, e.g., a mobile device, a wearable device, or an implantable device. The stimulus may be a sound signal, comprising at least one of a predetermined soundtrack, a tone, and white noise. The sound signal may comprise a soundtrack representing at least one of a sound of rainfall, a sound of a waterfall, a sound of ocean waves, a lullaby, a melody, and a polyphony.

The effect of the stimulus may be monitored by feedback, e.g., EEG, body temperature, heart rate, respiration rate, facial expression, muscle tone, vasodilation, which may be measured by non-contact sensors or wearable devices, and other electronic sensors embedded in the bed, blanket, mattress, sheets, pillow, etc. Body movement and eye movement may be observed by a video camera or webcam. The sensor signals are advantageously transmitted back to the user device to adjust the regime of stimulation. Of course, the communication path may be indirect to the user device, or the analysis of the signals may be remote from the user device, e.g., in a cloud computing center. An important aspect of the system is synchronizing the cycles with the context and current state of the subject. For example, if the subject got up to go to the bathroom or woke up for other reasons, the modulation cycle would generally need to restart from sleep stage 1. However, depending on the mental state of the subject, the progression through the sleep states may be varied.

The sound may be amplitude modulated on a carrier waveform, which would generally have higher frequencies that the modulation waveform (typically <100 Hz), and/or frequency modulated. When the sound separation between ears is present, the amplitude, frequency, phase, timing, etc. between ears may be modulated. Similarly, optical signals may be modulated by intensity, color, frequency, phase, etc., in addition to morphological objects and dynamic changes in real time.

The at least one waveform may be derived from an EEG recordings of brainwaves of at least one sleeping donor, processed using at least one of a principal component analysis (PCA), a correspondence analysis (CA), a factor analysis, a K-means clustering, a non-negative matrix factorization (NMF), a sparse PEA, a non-linear PCA a robust PCA, an independent component analysis (ICA), a network component analysis, and a singular spectral analysis.

Gender of the subject may be determined, and the gender of the subject marched with a gender of said at least one sleeping donor.

The at least one specific stage of sleep may be one of stage 1 of sleep, stage 2 of sleep, stage 3 of sleep, and stage 4 of sleep. In some analyses, 4 different non-REM (NREM) stages are classified, with stages 3 and 4 being deep sleep stages. See, nu.sleepassociation.org/about-sleep/stages-of-sleep/, FIG. 17. At least one specific stage of sleep may be one of REM sleep, non-REM sleep, and slow-wave sleep. At least one specific stage of sleep may be at least one complete sleep cycle comprising a natural sequence of sleep stages from stage 1 through stage 4 (REM). The at least one complete sleep cycle may comprise least three sequential complete sleep cycles.

The user device may comprise at least one speaker and wherein the stimulus comprises a sound signal delivered through said at least one speaker, and comprises an isochronic tone. The sound signal may be delivered to the subject through a pair of wireless earbuds, e.g., the modulated selected waveform may comprise binaural beats.

The user device may be configured to control an ambient light, which is selectively controllable to change at least one of brightness and color, and wherein the stimulus comprises a light signal which is presented to the subject through the ambient light. The light signal may be generated by at least one light emitting diode (LED). The LED may be disposed in proximity to the subjects eyes, e.g., in a sleep mask.

The user device may comprise at least one biometric sensor, further comprising the step of monitoring and collecting biometric data of the subject from said at least one biometric sensor.

The method may further comprise monitoring movement of the subject using at least one of a camera in the user device and a webcam coupled with the user device, processed with a neural network configured to classify a subjects sleep as one of a REM sleep, non-REM sleep, and a slow-wave sleep; and adjusting the stimulating of the subject upon determining whether the classification.

The method may further comprise monitoring a facial expression of the subject to determine if the subject is asleep or awake, and controlling a sequence of sleep stages induced by said stimulating in dependence on at least the monitored facial expression. The stimulating may be controlled to progress according to a natural series of sleep stages, further comprising resetting the progress according to the natural series of sleep stages in dependence on an awakening of the subject determined based on the monitored facial expression. The facial expression may be monitored by at least one of a camera in the user device and a webcam communicating with the user device. The facial expression may be monitored according to a signal present in at least one electromyographic electrode.

The method may further comprise obtaining biofeedback from the subject in real time and adjusting the stimulation of the subject in real time according to a biofeedback loop implemented by the user device.

It is also an object to provide a mobile device, comprising a housing; a microprocessor disposed within the housing; and a non-volatile memory disposed within the housing and electrically coupled with the processor, configured to store at least one app for controlling the microprocessor; the at least one app being configured to: (a) select a waveform from a plurality of waveforms derived from brainwaves of at least one sleeping donor, wherein said waveform corresponds to at least one specific stage of sleep; and (b) stimulate a subject with said at least one stimulus, wherein at least one stimulus selected from the group consisting of at least one of an auditory stimulus and a visual stimulus is modulated with the selected waveform, to thereby entrain the brain of the subject with the selected waveform to facilitate sleep in the subject. The mobile device may further comprise a battery electrically coupled with the processor, a display, disposed within the housing, electrically coupled with the microprocessor, a wireless communication transceiver disposed within the housing, electrically coupled with the microprocessor, at least one microphone, electrically coupled with the processor, at least one speaker disposed within the housing, electrically coupled with the processor, and at least one camera electrically coupled with the processor. The mobile device may be wirelessly coupled with a wearable device, wherein said wearable device comprises at least one biometric sensor configured to communicate biometric data from the subject to the mobile device through the wireless communication transceiver. The housing may be wearable by the subject and/or maintained close to the skull of the subject with a headband.

Another object provides a method of brain entrainment to facilitate sleep in a subject using a sleep app, comprising opening the sleep app on a programmable device; choosing at least one stimulus, wherein said at least one stimulus is one of a light signal and a sound signal; choosing a waveform from a menu having a plurality of waveforms derived from brainwaves of at least one sleeping donor, wherein said waveform corresponds to at least one specific stage of sleep; and stimulating the subjects brain with said at least one stimulus, wherein said at least one stimulus is modulated with the chosen waveform to entrain the brain of the subject with frequencies of the brainwaves of the at least one sleeping donor, to facilitate sleep in the subject. The method may further comprise recording a subjects electroencephalogram (EEG) during sleep while stimulated; and adjusting the stimulating based on the subjects electroencephalogram (EEG) in real time using a neurofeedback loop.

A further object provides a method of brain entrainment to facilitate sleep in a subject, comprising: providing a programmable device having a sleep app stored in a non-volatile memory; providing a stimulator, selected from one or more of a light stimulator and a sound stimulator, defining a waveform, by the sleep app, from a plurality of waveforms, each respective waveform being derived from brainwaves of at least one sleeping donor, wherein said waveform corresponds to at least one specific stage of sleep; stimulating the subject with the stimulator, having at least one of a light output or a sound output modulated with the defined waveform, to entrain the brain of the subject with the brainwaves of the at least one sleeping donor, to facilitate sleep in the subject. The method may further comprise recording an electroencephalogram (EEG) from the subject during sleep while being stimulated, and defining at least one new waveform for stimulation, said waveform being selectively dependent on the uploaded recorded electroencephalogram. The at least one new waveform may be used to stimulate the subject one day after the electroencephalogram is recorded. The recorded electroencephalogram may be uploaded to a remote server, and the new waveform for stimulation subsequently downloaded from the remote server.

Another object provides a non-transitory computer readable medium storing instructions for controlling a processor to perform a method comprising: instructions to select a waveform from a plurality of waveforms derived from brainwaves of at least one sleeping donor, wherein said waveform corresponds to at least one specific stage of sleep; and instructions to stimulate a subject with said at least one stimulus, wherein at least one stimulus selected from the group consisting of at least one of an auditory stimulus and a visual stimulus is modulated with the selected waveform, to thereby entrain the brain of the subject with the selected waveform to facilitate sleep in the subject.

A still further object provides a method of generating a waveform for neuromodulation to improve sleep in a subject, the method comprising: collecting EEG recording from at least one sleeping donor; identifying portions of the EEG recordings corresponding to a specific sleep stage; grouping a plurality of portions of the EEG recordings corresponding to the specific sleep stage, each group corresponding to the specific sleep stage; analyzing each group corresponding to the specific sleep stage using a statistical analysis; extracting a cortical signature corresponding to each specific sleep stage; generating a waveform based on the cortical signature for each sleep stage; and modulating a stimulus for the subject according to the waveform. The modulating of the stimulus may be performed under control of a sleep app executing on a mobile or wearable device. The statistical analysis may be at least one of a principal component analysis (PCA), a correspondence analysis (CA), a factor analysis, a K-means clustering, a non-negative matrix factorization (NMF), a sparse PEA a non-linear PCA a robust PEA an independent component analysis (ICA), a network component analysis, and a singular spectral analysis.

It is, therefore, an object to provide a method of inducing sleep in a second subject comprising: recording brain activity patterns of a first subject (donor) who is asleep; and inducing sleep in the second subject (recipient) by replicating the brain activity patterns of the donor in the recipient.

It is also an object to provide a method of preventing sleep in a second subject (recipient) comprising: recording brain activity patterns of a first subject (donor) who is awake; and preventing sleep in the second subject (recipient) by replicating the brain activity patterns of the donor in the recipient.

It is further an object to provide a method of inducing sleep in a second subject (recipient) comprising: identifying the mental state of a first subject (donor); if the donor is asleep, recording brain activity patterns of the donor, and inducing sleep in the recipient by replicating the brain activity patterns of the donor in the recipient. The method may further comprise verifying that the recipient is asleep.

It is still a further object to provide a method of preventing sleep in a second subject (recipient) comprising: identifying a mental state of a first subject (donor); if the donor is awake, recording brain activity patterns of the first subject; and preventing sleep in the second subject by replicating the brain activity patterns of the second subject. The method may further comprise verifying that the second subject is awake.

Another object is a method of transplanting a desired mental state from a first subject (donor) to a second subject (recipient) comprising: identifying a mental state of the donor, capturing a mental state of the donor by recording brain activity patterns; saving the brain activity patterns in a non-volatile memory; retrieving the brain activity patterns from the non-volatile memory; and transplanting the desired mental state of the donor to the recipient by inducing the brain activity patterns in the recipient, wherein the desired mental state is one a sleeping state and a waking state.

Another object is a method of transplanting a desired sleep stage from a first subject (donor) to a second subject (recipient) comprising: identifying a sleep stage of the donor, capturing a sleep stage of the donor by recording brain activity patterns; saving the brain activity patterns in a non-volatile memory; retrieving the brain activity patterns from the non-volatile memory; and transplanting the desired sleep stage of the donor to the recipient by inducing the brain activity patterns in the recipient, wherein the desired sleep stage is one a sleep stage 1, 2, and 3.

Another object is a method of transplanting a desired sleep stage from a first subject (donor) to a second subject (recipient) comprising: identifying a sleep stage of the donor, capturing a sleep stage of the donor by recording brain activity patterns; saving the brain activity patterns in a non-volatile memory; retrieving the brain activity patterns from the non-volatile memory; and transplanting the desired sleep stage of the donor to the recipient by inducing the brain activity patterns in the recipient, wherein the desired sleep stage is one of a REM sleep stage and non-REM sleep stage.

Another object is a method of transplanting a desired sleep stage from a first subject (donor) to a second subject (recipient) comprising: identifying a sleep stage of the donor, capturing a sleep stage of the donor by recording brain activity patterns; saving the brain activity patterns in a non-volatile memory; retrieving the brain activity patterns from the non-volatile memory; and transplanting the desired sleep stage of the donor to the recipient by inducing the brain activity patterns in the recipient, wherein the desired sleep stage is a slow-wave deep non-REM sleep.

A further object is a method of improving sleep in a recipient by transplanting a mental state of a donor to the recipient comprising: recording brainwaves of the donor, and transplanting the mental state of the donor to the recipient by inducing the recorded brainwaves of the donor in the recipient, wherein the mental state is one of a waking state and a sleeping state.

A still further object is a method of transplanting a desired mental state of a first subject (donor) to a second subject comprising: identifying a mental state of the donor, recording brainwaves of the donor in a desired mental state; and transplanting the desired mental state of the donor to the recipient by inducing the brainwaves of the first subject in the second subject, wherein the desired mental state is one of a sleeping state and a waking state.

Another object is a method of improving sleep in a recipient by transplanting the desired state of a healthy sleep to the recipient comprising: identifying a mental state of the plurality of healthy donors; recording brainwaves of the plurality of healthy donor in a state of sleep; saving the brainwaves in a non-volatile memory; retrieving the brainwaves from the non-volatile memory; and transplanting the state of healthy sleep from the plurality of healthy donors to the recipient by inducing the brainwaves of the donor in the recipient. The method may comprise identifying a mental state of the recipient to verify that the recipient has the desired mental state. The brainwaves may be recorded using EEG, qEEG, or MEG. The method may further comprise filtering the recorded brainwaves from noise and/or performing PCA to determine dominant frequencies and secondary (and, possibly, higher) harmonics.

A further object is a system for transplanting a desired mental state from a first subject (donor) to a second subject (recipient) comprising: a first apparatus for recording brainwaves of the donor in a desired mental state; a non-volatile memory coupled with the first apparatus for storing the recording of the brainwaves; and a second apparatus for inducing the brainwaves in the recipient to transplant to the recipient the desired mental state of the donor, the second apparatus configured to receive the recording of the brainwaves of the donor from the non-volatile memory, wherein the desired mental state is one of a sleeping state and a waking state. The first apparatus may be one of an electroencephalograph and a magnetoencephalograph. The second apparatus may be one of a source of light signal or sound signal configured to modulate donors brainwave frequencies on the light signal or the sound signal.

Another object is a method of transplanting a desired mental state of a first subject (donor) to a second subject (recipient) comprising: identifying a mental state of the donor, recording at least one of EEG and MEG of the donor, said donor being in a desired mental state; processing the EEG or MEG signal; saving the processed signal in a nonvolatile memory; retrieving the processed signal from the nonvolatile memory; modulating the processed signal on at least one stimulus; and transplanting the desired mental state of the first subject to the second subject by stimulating the second subject with said at least one stimulus, wherein the desired mental state is a sleeping state or a waking state. The processing may comprise removing noise from the EEG or MEG signal; and/or compressing the EEG or MEG signal. The EEG or MEG signal retrieved from the nonvolatile memory may be decompressed. The stimulus may be a light signal, a sound signal, or a combination thereof. The light stimulation may be an ambient light or a direct light. The sound stimulation may be binaural beats or isochronic tones.

A still another object is a system for transplanting a desired mental state of a first subject (donor) to a second subject (recipient) comprising: an electroencephalograph or a magnetoencephalograph for recoding brainwaves of the donor, the donor being in a desired mental state; a processor coupled with an electroencephalograph or a magnetoencephalograph, the processor configured to perform signal analysis and calculate at least one dominant frequency of the brainwaves of the donor, a nonvolatile memory coupled with the first processor for storing the at least one frequency of the brainwaves of the donor, a second apparatus for inducing the brainwaves in the recipient to transplant to the recipient the desired mental state of the donor, the second apparatus configured to receive said at least one dominant frequency of the brainwaves of the donor from the non-volatile memory, wherein the desired mental state is one of a sleeping state and a waking state.

The second apparatus may be a light source capable of modulating said at least one dominant frequency on the light, a sound source capable of modulating said at least one dominant frequency on the sound, or a combination thereof. The sound source may be binaural beats source or isochronic tones source.

A further object is a method of transplanting a circadian rhythm of a first subject (donor) to a second subject (recipient) comprising: recording EEG or MEG of the donor, the donor having a desirable phase of the circadian rhythm; processing the recorded EEG or MEG to remove noise; saving the processed EEG or MEG in a nonvolatile memory; retrieving the processed EEG or MEG from the nonvolatile memory; and transplanting the desired phase of the circadian rhythm of the donor to the recipient by “playing back” the processed EEG or MEG of the donor to the recipient via sensory stimulation or other one or more stimulus on which the donors EEG or MEG is modulated. The method may further comprise compressing the recorded EEG or MEG, before saving it in the non-volatile memory; and decompressing the recorded EEG or MEG after retrieving compressed EEG or MEG from the non-volatile memory.

Yet another object is a system for transplanting a circadian rhythm of a first subject (donor) to a second subject (recipient) comprising: an electroencephalograph or a magnetoencephalograph for recording EEG or MEG respectively; a first processor coupled to the electroencephalograph or the magnetoencephalograph and configured for digital signal processing for removing noise from the recorded EEG or MEG; a non-volatile memory coupled with the processor for storing the processed EEG or MEG; and a stimulation device coupled to the non-volatile memory for playing back the processed EEG or MEG to the recipient to induce the circadian rhythm of the donor to the recipient. The stimulation device may be a sensory stimulation device, a source of light or a source of the sound, each capable of modulating recorded EEG or MEG on a light signal or a sound signal respectively. The first processor may be further configured to compress the processed EEG or MEG. A second processor configured to decompress compressed EEG or MEG may be coupled to the non-volatile memory and to the stimulation device.

The technology may be used to modify or alter a mental state (e.g., from sleep to waking and vice versa) in a subject. Typically, the staging mental state, brain state, or brainwave pattern is assessed, such as by EEG, MEG, observation, stimulus-response amplitude and/or delay, or the like. Of particular interest in uncontrolled environments are automated mental state assessments, which do not rely on human observation or EEG signals, and rather may be acquired through MEG (e.g., SQID, optically-pumped magnetometer), EMG, MMG (magnetomyogram), mechanical (e.g., accelerometer, gyroscope, etc.), data from physiological sensors (e.g., EKG, heartrate, respiration rate, temperature, galvanic skim potential, etc.), or automated camera sensors.

For example, cortical stimulus-response pathways and reflexes may be exercised automatically, to determine their characteristics on a generally continuous basis. These characteristics may include, for example, a delay between stimulus and the observed central (e.g., EEG) or peripheral response (e.g., EMG, limb accelerometer, video). Typically, the same modality will be used to assess the pre-stimulation state, stimulus response, and post-stimulation state, though this is not a limitation.

In order to change the mental state, a stimulus is applied in a way designed to alter the mental state in the desired manner. A state transition table, or algorithm, may be employed to optimize the transition from a starting mental state to a desired mental state. The stimulus may be provided in an open loop (predetermined stimulus protocol) or closed loop (feedback adapted stimulus protocol), based on observed changes in a monitored variable.

Advantageously, a characteristic delay between application of stimulus and determination of response varies with the brain or mental state. For example, some mental states may lead to an increased delay or greater variability in delay, while others may lead to decreased or lower variability. Further, some states may lead to attenuation of response, while others may lead to an exaggerated response. In addition, different mental states can be associated with qualitatively different responses. Typically, the mere assessment of the brain or mental state should not itself alter the state, though in some cases the assessment and transition influence may be combined. For example, in seeking to assist in achieving a deep sleep state, the excitation that disturbs sleep is contraindicated.

In cases where a brainwave pattern is itself determined by EEG (which may be limited to relatively controlled environments), brainwaves representing that pattern represent coherent firing of an ensemble of neurons, defining a phase. One way to change the state is to advance or retard the triggering of the neuronal excitation, which can be a director indirect excitation or inhibition, caused, for example, by electrical, magnetic mechanical, or sensory stimulation. This stimulation may be time-synchronized with the detected (e.g., by EEG) brainwaves, for example with a phase lead or lag with respect to the detected pattern. Further, the excitation can steer the brainwave signal by continually advancing to the desired state, which through the continual phase rotation represents a different frequency. After the desired new state is achieved, the stimulus may cease, or be maintained in a phase-locked manner to hold the desired state.

A predictive model may be used to determine the current mental state, optimal transition to a desired mental state, when the subject has achieved the desired mental state, and how to maintain the desired mental state. The desired mental state itself may represent a dynamic sequence (e.g., stage 1→stage 2→stage 3, etc.), such that the subjects mental state is held for the desired period in a defined condition. Accordingly, the stimulus may be time-synchronized with respect to the measured brainwave pattern.

Direct measurement or determination of brainwaves or their phase relationships is not necessarily required. Rather, the system may determine tremor or reflex patterns. Typically, the reflex patterns of interest involve central pathways, and more preferably brain reflex pathways, and not spinal cord mediated reflexes, which are less dependent on instantaneous brain state. The central reflex patterns can reflect a time delay between stimulation and motor response, an amplitude of motor response, distribution of response through various afferent pathways, the variability of response, tremor or other modulation of motor activity, etc. Combinations of these characteristics may be employed, and different subsets may be employed at different times or to reflect different states. Similar to evoked potentials, the stimulus may be any sense, especially sight, sound, touch/proprioception/pain/etc., though the other senses, such as taste, smell, balance, etc., may also be exercised. A direct electrical or magnetic excitation is also possible. As discussed, the response may be determined through EEG, MEG, or peripheral afferent pathways.

A further object provides a system and method for enhancing deep non-REM deep, comprising statistically separating slow-wave sleep components from acquired brainwave patterns; defining a stimulation pattern based on the statistically separating slow-wave sleep components, and stimulating a subject with the defined stimulation pattern. The neurological stimulator comprises a memory configured to store acquired brainwave patterns; at least one processor configured to: statistically separate slow-wave non-REM sleep components from the acquired brainwave patterns; and define a brain stimulation pattern based on the statistically separating slow-wave non-REM deep sleep components; and an output signal generator configured to defined brain stimulation pattern.

A still further object provides a system and method for enhancing deep sleep, comprising: extracting brainwave patterns representing a deep sleep state comprising slow wave sleep, from indigenous brain activity EEG recordings of at least one subject; processing the extracted brainwave patterns using a statistical processing algorithm to separate slow wave sleep components from the indigenous brain activity EEG recordings of the at least one subject; inverting the processed extracted brainwave patterns; and stimulating a subject with the inverted processed extracted brainwave patterns. The corresponding system for enhancing deep sleep comprises a memory configured to store brainwave patterns representing a deep sleep state comprising slow wave sleep, from indigenous brain activity EEG recordings of at least one subject at least one processor configured to process the extracted brainwave patterns using a statistical processing algorithm to separate slow wave sleep components from the indigenous brain activity EEG recordings of the at least one subject; and a stimulator, configured to generate a stimulation signal based on the processed extracted brainwave patterns. The stimulator may comprise a sensory stimulator (e.g., sight, sound, vestibular, touch, taste, smell, etc.). In order to format the signal for stimulating the brain, it may be inverted. Normalization of brain activity information may be spatial and/or temporal.

See also US 2016/0066838 (DeCharms); US 2009/0099623 (Bentwich); US 201210289869 A1 (Tyler); US 2004/0131998 (Marmon et al.); U.S. Pat. No. 5,356,368 (Monroe); US 2002/0198577 (Jaillet); and US 2015/0291074 (Advanced Telecommunications Research Institute International). DeCharms discloses a computer-assisted method for treating pain in a subject comprising measuring activity of one or more internal voxels of a brain of said subject associated with pain; communicating instructions to said subject which modulate the activity of said voxel, and training said subject to control said internal voxel. DeCharms provides methods, software, and systems that may be used to provide and enhance the activation and control of one or more regions of interest, particularly through training and exercising those regions of interest. Data analysis/behavioral control software performs computations of brain scan image data to produce activity metrics that are measures of physiological activity in brain regions of interest. The results and other information and ongoing collected data may be stored to data files of progress and a record of the stimuli used. The selected instruction, measured information, or stimulus, is then presented via a display to a subject. This encourages the subject to engage in imagined or performed behaviors or exercises or to perceive stimuli. If the subject undertakes overt behaviors, such as responding to questions, the responses and other behavioral measurements are fed to the data analysis/behavioral control software. According to DeCharms, a subject can be trained to control the activation of a region of interest of that subject's brain, and then exercise the use of that region to further increase the strength and control of its activation. This training and exercise can have beneficial effects on the subject. In the case of regions that release endogenous neuromodulatory agents, this control can serve a role similar to that of externally applied drugs.

It is also an object to provide a method of generating a waveform for neuromodulation to improve sleep in a subject, the method comprising: collecting EEG recordings from at least one sleeping donor for a plurality of sleep stages; grouping a plurality of portions of the EEG recordings corresponding to the plurality of sleep stages, into a plurality of groups corresponding to the plurality of sleep stages; analyzing each group using a statistical analysis; extracting a cortical signature corresponding characteristic of each analyzed group; generating a waveform based on the characteristic comical signature for each of the plurality of sleep stages; and modulating a stimulus for the subject according to the generated waveforms for the plurality of sleep stages.

It is a further object to provide a mobile device contained within a housing, comprising: a microprocessor; an electrical power source, electrically coupled with the microprocessor; a wireless communication transceiver, electrically coupled with the microprocessor, at least one microphone port, electrically coupled with the microprocessor, configured to receive an electrical signal corresponding to a sound; at least one camera port electrically coupled with the microprocessor, configured to receive an electrical signal corresponding to an image; a display, electrically coupled with the microprocessor, at least one speaker port, electrically coupled with the microprocessor, configured to generate an electrical signal corresponding to a sound; a non-volatile memory and electrically coupled with the microprocessor, configured to store at least one app downloadable through the wireless communication transceiver for controlling the microprocessor, said at least one downloadable app being configured to: (a) select a waveform from a plurality of waveforms derived from brainwaves of at least one sleeping donor, wherein said waveform corresponds to at least one a specific stage of sleep, a gender, and an age group; and (b) define a stimulus for stimulation of a subject, selected from the group consisting of at least one of an auditory stimulus generated through the at least one speaker, and a visual stimulus generated through the display, modulated with the selected waveform, and adapted to entrain the brain of the subject with the selected waveform to facilitate sleep in the subject; wherein at least one of the selection of the waveform and the definition of the stilulus is responsive to the at least one microphone or the at least one camera.

It is another object to provide a method of facilitating sleep, comprising: providing data defining a plurality of waveforms in a memory; retrieving a selected waveform from the memory, selectively dependent on at least one of a determined sleep phase of a human subject and a predetermined sequence; and stimulating the human subject with a stimulus modulated according to the selected waveform; to thereby entrain the brain of the human subject with the selected waveform to facilitate sleep in the subject.

The plurality of waveforms in the memory may be derived from brain activity measurements acquired during at least one sleep cycle of at least one human, or from brain activity measurements acquired during at least one sleep cycle of the human subject.

The method may further comprise acquiring brain activity measurements during at least one sleep cycle of at least one human; and processing the acquired brain activity measurements to define the plurality of waveforms in the memory.

The stimulus may be modulated in a human subject device, according to a sleep app stored within the human subject device, the sleep app being downloadable and upgradeable from a remote server.

The predetermined sequence may be defined by a human user interface menu of a human subject device for selecting at least one respective waveform.

The sleep phase of the human subject may be determined based on at least electroencephalographic activity of the human subject or based on at least bioelectric signals received from the human subject.

The stimulus modulated according to the selected waveform may entrain the brain of the human subject with the selected waveform to facilitate sleep in the human subject.

The stimulus modulated according to the selected waveform may be one of alight stimulus and a sound stimulus.

The selected waveform may correspond to at least one specific stage of sleep.

Each of the plurality of waveforms may be derived from recordings of brainwaves of at least one sleeping donor, processed using a statistical decision analysis.

The method may further comprise adaptively defining a sequence of sleep stages dependent on biometric information received from a sleeping human subject; and selecting waveforms from the memory in dependence on a correspondence to a respective sleep stage of the adaptively defined sequence of sleep stages; wherein said stimulating the human subject comprises altering a sleep stage of the human subject dependent on both the determined sleep phase of a human subject and the adaptively defined sequence of sleep stages.

The human subject may be stimulated with at least one audio transducer and wherein the stimulus comprises at least one of an isochronic tone and binaural beats or with an ambient light stimulus, selectively modulated according to the selected waveform to change at least one of brightness and color. The ambient light stimulus may be emitted by at least one light emitting diode disposed in a sleep mask proximate the human subjects eyes.

The method may further comprise providing at least one sensor to determine at least one of an eye movement and a facial expression of the human subject, to at least one of determine a current determined sleep phase of a human subject or select the predetermined sequence.

The predetermined sequence may be a natural series of sleep stages, the method further comprising resetting the progress according to the natural series of sleep stages in dependence on an awakening of the human subject.

BRIEF DESCRIPTION OF THE DRAWINGS

The detailed description is described with reference to the accompanying figures. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The use of the same reference number in different figures indicates similar or identical items.

FIG. 1 shows a flowchart according to one embodiment of the invention illustrating a process of replicating a sleep state from one subject to another subject.

FIG. 2 shows a flowchart according to one embodiment of the invention illustrating a process of replicating a waking stage from one subject to another subject by recording and replicating brainwaves associated with the waking stage, according to one embodiment of the invention.

FIG. 3 shows a flowchart according to one embodiment of the invention illustrating a process of replicating a sleep stage from at least one first subject to another subject by recording electroencephalogram (EEG) of said least one first subject, extracting at least one dominant frequency from the EEG and replicating the sleep stage of said at least one first subject in a second subject by stimulating the second subject with stimuli having the dominant frequency associated with the desired sleep stage, according to one embodiment of the invention.

FIG. 4 shows a flowchart according to one embodiment of the invention illustrating a method of improving sleep in a recipient by recording EEG or MEG of a healthy donor and “playing it back” to the recipient via transcranial stimulation.

FIG. 5 shows a flowchart according to one embodiment of the invention illustrating the creation of a database of sleep stages and their associated frequencies for later brain entrainment.

FIG. 6 shows a flowchart according to one embodiment of the invention illustrating using a neural network in the creation of a database of sleep stages and their associated frequencies for later brain entrainment.

FIG. 7 shows a flowchart according to one embodiment of the invention illustrating a method of recording a mental state of a first subject in a desirable state of the subject's circadian rhythm and transplanting this mental state into another subject to replicated the desirable state of the circadian rhythm.

FIG. 8 shows a flowchart according to a further embodiment of the invention.

FIG. 9 shows a flowchart according to one embodiment of the invention illustrating a process of replicating the desired sleep stage from one subject to another subject.

FIG. 10 shows a flowchart according to an embodiment of the invention.

FIG. 11 shows a flowchart according to an embodiment of the invention.

FIG. 12 shows a flowchart according to an embodiment of the invention.

FIG. 13 shows a flowchart according to an embodiment of the invention.

FIG. 14 shows a flowchart according to an embodiment of the invention.

FIG. 15 shows a flowchart according to an embodiment of the invention.

FIG. 16 shows a schematic representation of a smartphone for executing apps.

FIG. 17 shows a hypnogram of a healthy adult.

FIG. 18 shows a hypnogram of a healthy adult.

FIG. 19 shows a sequence of sleep stages in a healthy adult.

FIG. 20A shows an original EEG recording of a REM phase in a 34 years old female.

FIG. 20B shows an EEG recording of a REM phase in a 34 years old female reconstructed with 64 SSA groups.

FIG. 20C shows an EEG recording of a REM phase in a 34 years old female reconstructed with 16 SSA groups.

FIGS. 21A and 21B show an EEG for a 30 years old female in sleep stage R.

FIG. 22 show an EEG for a 30 years old female in sleep stage 3.

FIGS. 23A and 23B show an EEG for a 30 years old female in sleep stage 3.

FIGS. 24A and 24B show an EEG for a 25 years old female in sleep stage W.

FIGS. 25A and 25B show an EEG for a 25 years old male in sleep stage 2.

FIGS. 26A and 26B show an EEG for a 25 years old male in sleep stage 1.

FIGS. 27A and 27B show an EEG for a 25 years old male in sleep stage W.

FIG. 28 shows a schematic diagram of a mental state transfer system.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that the present disclosure may be readily implemented by those skilled in the art. However, it is to be noted that the present disclosure is not limited to the embodiments but can be embodied in various other ways. In drawings, parts irrelevant to the description are omitted for the simplicity of explanation, and like reference numerals denote like parts through the whole document.

FIG. 1 shows a flowchart of a first embodiment according to the present invention. A first subject (donor), having a mental state, is interrogated, observed or sensed, to determine or identify his or her mental state 100. The first subject is typically human, though this is not a limit of the technology and the subject may be an animal. In this embodiment, the process seeks to identify a characteristic sleep pattern, and therefore the mental state of the first subject is monitored until a sleep state occurs 110. When the first subject (donor) is asleep, brain activity patterns reflecting or characterizing the sleep state are captured 120. This step may be done by recording EEG or MEG of the first subject (donor). And the brain activity patterns are stored in a non-volatile memory 130. These stored patterns may be optionally processed, statistically aggregated, analyzed for perturbations or anomalies, filtered, compressed, etc. Stages of sleep may be determined. It is noted that brain activity patterns change over time during sleep from stage to stage, and therefore, the stored patterns may encompass one or more stages of sleep.

The stored data from the first subject (donor) is then used to induce sleep in a second subject (a recipient—also typically a human, but may be an animal) by replicating the brain activity patterns (or sequences of brain activity patterns) of the first subject (donor) in the second subject (recipient) 150. The replication of brain activity patterns, dependent on the stored patterns, typically seeks to stimulate or induce the brain of the second subject (recipient) by modulating a stimulus (or several stimuli) in a manner synchronized with the frequency, phase and/or waveform pattern represented in the brain activity patterns of the first subject (donor) in the sleep state. Typically, when the second subject (recipient) achieves the sleep state 160 (assuming that the first subject and second subject are physiologically compatible—a donor and a recipient should both be either human or animals), the brain activity patterns of the first and second subject will be corresponding.

According to the present technology, the modulation of stimulation, which is, for example, a sensory stimulation, whose waveform is modulated to correspond to the raw or processed brainwave pattern of the first subject (donor) for the brain region associated with the stimulation electrode.

For example, the brain activity pattern of the first subject (donor) is measured by EEG electrodes. In a sleep state, it may assume various wave patterns, over the range <1 Hz to about 25 Hz, which vary in amplitude, frequency, spatial location, and relative phase. For example, the first stage of sleep is initially dominated by alpha brainwaves with a frequency of 8 Hz to 13 Hz. Typically, brain activity pattern measurement from the first subject (donor) has a higher spatial resolution, e.g., 64 or 128 electrode EEGs, than the stimulator for the second subject (recipient), and the stimulus electrodes tend to be larger than the EEG electrode. The stimulus for the second subject (recipient) is therefore processed using a dimensionality (or spatial) reduction algorithm to account for these differences, which will tend to filter the stimulus signal. By applying this stimulus modulated with the brain activity of the first subject (donor), the second subject (recipient) is made susceptible to synchronization with the brain activity pattern of the first subject (donor). For example, by temporally modulating the polarization level of the cells near the electrode, the cells will better couple to excitation stimuli in the brain of the second subject (recipient) having the characteristics of the brain activity pattern of the first subject (donor).

The donors indigenous brainwaves may be modulated on light, sound, vibrations or any number of other stimuli amenable to frequency modulation. For example, donors brainwaves may be modulated on ambient light, on binaural beats, or isochronic tones.

The verification that the recipient has achieved the desired sleep state may optionally be done by visual observation, by EEG, EKG, measuring heart and/or respiration rate, body temperature or any number of other physiological parameters that will be well understood by a person skilled in the art. These measurements should be, preferably, done automatically via biosensors.

FIG. 2 shows a flowchart of the second embodiment according to the present invention. A first subject (donor), having a mental state, is interrogated, observed or sensed, to determine or identify of his or her mental state 100. The first subject is typically human, though this is not a limit of the invention (which equally applies to any animal). In this embodiment, the interrogation seeks to identify a characteristic alert/awake pattern, and therefore the mental state of the first subject is monitored until an alert state occurs 111. When the first subject (donor) is awake, brain activity patterns reflecting or characterizing the waking state are captured 120, and stored in a non-volatile memory 130. For example, one may seek to capture the patterns that represent awakening, and therefore, the monitoring commences on a sleeping subject. These stored patterns may be optionally processed, statistically aggregated, analyzed for perturbations or anomalies, filtered, compressed, etc Stages of awakening may be determined. It is noted that the brain activity patterns change over time during awakening, and therefore, the stored patterns may encompass one or more stages of the waking process.

The stored data from the first subject (donor) is then retrieved from the non-volatile memory 140 and used to “transplant” the state of alertness to prevent sleep, or maintain alertness, in a second subject (a recipient—also typically, but not necessarily, a human) by replicating the awake brain activity patterns of the first subject (donor), or sequences of brain activity patterns, in the second subject (recipient) 170. The replication of brain activity patterns, dependent on the stored patterns, typically seeks to stimulate or induce the brain of the second subject (recipient) by modulating indigenous brainwaves of the donor on a stimulus in a manner synchronized with the frequency, and preferably phase and/or waveform pattern represented in the brain activity patterns of the first subject (donor) in the awake or wakening state. Typically, when the second subject is awake or wakes up, 180, the brain activity patterns of the first and second subject will be corresponding.

FIG. 3 shows a flowchart of a third embodiment, in which the technology is generalized. A first subject (donor), having a mental state, is interrogated, observed or sensed, to determine or identify his or her mental state 190. The mental state of the first subject is monitored until the desired state is achieved 200. When the first subject achieves that state, brain activity patterns reflecting or characterizing the state are captured 210 by, for example, recording EEG or MEG of the first subject, and optionally stored in non-volatile memory. The brain activity pattern is, e.g., brainwaves (e.g., EEG) 210.

The brainwaves are analyzed using statistical data mining techniques such as principal component analysis (PCA) to determine a set of linearly-uncorrelated variables—principal components. At least one dominant frequency in the recorded brainwaves is identified 220. Optionally, secondary and higher harmonics may be identified as well. It will be well-understood by a person skilled in the ad that any number of similar statistical data analysis techniques may be used, such as signal processing, independent component analysis, network component analysis, correspondence analysis, multiple correspondence analysis, factor analysis, canonical correlation, functional principal component analysis, independent component analysis, singular spectrum analysis, weighted PEA, sparse PCA principal geodesic analysis, eigenvector-based multivariate analyses, etc.

The stored data from the first subject is then retrieved, at least the dominant frequency is modulated on at least one stimulus and used to “transplant” the desired mental state of the donor in a second subject (recipient) by seeking to replicate the brain activity patterns of the first subject (donor), or sequences of brain activity patterns, in the second subject (recipient) 240. The second subject (recipient) is then monitored for induction of the desired mental state 250.

FIG. 4 shows a flowchart according to the fourth embodiment, in which an EEG or EMG of a first subject (healthy donor), while in a state of sleep, is recorded 260, optionally processed to remove noise 270, and stored 280. The data may optionally be compressed. The stored data is retrieved 290 and decompressed as necessary. The data is then played back to a second subject (recipient), to improve the quality of sleep 300.

FIG. 5 shows a flowchart according to the fifth embodiment, in which a multichannel EEG/EMG of a first subject (donor) is recorded 310, and processed to remove noise (and/or artifacts) and/or compress the data 320. It is optionally stored in non-volatile memory. PCA analysis is performed on the data to determine characteristic frequencies associated with sleep stages 330. A database is created, storing the recorded EEG/MEG, the associated characteristic frequencies, and corresponding sleep stages, so that a characteristic frequency may be retrieved for any given sleep stage 340. This database can be a relational database or any other type of searchable database as will be readily understood by anyone skilled in the art. According to the sixth embodiment, a multichannel EEG/EMG of a first subject (donor) is recorded 310, and processed to remove noise (and/or artifacts) and/or compress the data 320. It is optionally stored in non-volatile memory. An artificial neural network is trained on this data to determine characteristic frequencies associated with sleep stages 350. A deep neural network, as well as other AI machine-learning tools, may be used as will be readily understood by a person skilled in the art. A database is created, storing the recording of the EEG/MEG, the associated characteristic frequencies, and corresponding sleep stages, so that a characteristic frequency may be retrieved for any given sleep stage 340.

FIG. 6 shows a flowchart according to an embodiment of the invention. A multichannel EEG or EMG of a plurality of healthy sleeping donors is recorded 310. The multichannel EEG/EMG recordings are processed too, e.g., remove noise 320. A neural network is trained on the recorded EEG/EMG recordings to identify characteristic frequencies associated with sleep stages 350. A database of sleep stage characteristic frequencies is created.

FIG. 7 shows a flowchart according to a further embodiment of the present invention illustrating a process in which a first subject (donor) is monitored with respect to phases of his or her circadian rhythm with his or her EEG or EMG recorded 360, processed to remove noise (and/or artifacts), and, optionally, compressed 270, and then stored in a non-volatile memory 280. In this case, the stored signals are tagged with the circadian cycle phase, unless only a single phase is captured, or pattern recognition used to identify the cycle stage. The stored data is then retrieved 290, decompressed 370, and played back to a second subject (recipient) 380, using sensory stimulation, or other stimuli, to induce a desired circadian rhythm state. In this case, the technology may also be used to prolong states in the second subject or hasten the transition from one state to another. It may also be used to treat circadian rhythm disorders, by reinforcing healthy or normal circadian rhythm patterns in a second subject with an otherwise abnormal cycle.

FIG. 8 shows a flowchart according to a further embodiment of the present invention illustrating a process of replicating the desired sleep stage from one subject (donor) to another subject (recipient). In general, the sleep stage of the source subject is determined in a traditional manner, which may include brain signal analysis, other biometrics, and/or observation. The data may be acquired 400 over one or more sleep cycles, and during or after different types of environmental conditions or stimulation. For example, various types of music may be played, seeking to entrain a conscious or subconscious rhythm. Lights can flash, and various other sensory stimulation may occur. The brain signal readings are synchronized and tagged with the stimulation parameters 410 so that the stimulation is associated with its respective effect. Similarly, before sleep, the subject may be presented with certain experiences, such that during sleep, the memory processing within the brain is dependent on these experiences.

After the various data is acquired from the subject 400, along with information about the pre-sleep experience and or context 410, and sensory Stimulation during sleep, a memory, database, statistical model, the rule-based model is generated, and/or neural network is trained, reflecting the subject (donor). Data may be aggregated from a plurality of subjects (donors), but typically, these are processed for the particular subject before aggregation. Based on single or multiple subject data, a normalization process may occur 420. The normalization may be spatial and/or temporal. For example, the EEG electrodes between sessions or for the different subject may be in different locations, leading to a distortion of the multichannel spatial arrangement. Further, the head size and shape of different individuals are different, and this needs to be normalized and/or encoded as well. The size and shape of the head/skull and/or brain may also lead to temporal differences in the signals, such as characteristic time delays, resonant or characteristic frequencies, etc.

One way to account for these effects is through the use of a time-space transform, such as a wavelet-type transform. It is noted that, in a corresponding way that statistical processes are subject to frequency decomposition analysis through Fourier transforms, they are also subject to time-frequency decomposition through wavelet transforms. Typically, the wavelet transform is a discrete wavelet transform (DINT), though more complex and less regular transforms may be employed. As discussed above, principal component analysis (PCA) and spatial PCA may be used to analyze signals, presuming linearity (linear superposition) and statistical independence of components. However, these presumptions technically do not apply to brainwave data, and practically, one would normally expect interaction between brain wave components (non-independence) and lack of linearity (since “neural networks” by their nature are non-linear), defeating the use of PCA or spatial PCA unmodified. However, a field of nonlinear dimensionality reduction provides various techniques to permit corresponding analyses under the presumptions of non-linearity and non-independence. See, en.wikipeda.org/wiki/Nonlinear_dimensionality_reduction, www.image.ucar.edu/pub/toyN/monahan_5_16.pdf (An Introduction to Nonlinear Principal Component Analysis, Adam Monahan), Nonlinear PCA toolbox for MATLAB (www.nlpca.org), Nonlinear PCA (www.comp.nus.edu.sg/cs5240/lecture/nonlinear-pca.pdf), Nonlinear Principal Components Analysis: Introduction and Application (openaccess.leidenuniv.nl/bitstream/handle/1887/12386/Chapter2.pdf?sequence=10, 2007), Nonlinear Principal Component Analysis: Neural Network Models and Applications (pdfs.semanticscholar.org/9d31/23542031a227d2f4c4602066cf8ebceaeb7a.pdf), Karl Friston, “Nonlinear PCA: characterizing interactions between modes of brain activity” (www.fil.ion.uctac.uk/karl/NonlinearPCA.pdf,2000), Howard et al., “Distinct Variation Pattern Discovery Using Alternating Nonlinear Principal Component Analysis”, IEEE Trans Neural Network Learn Syst 2018 January; 29(1):156-166. doi: 10.1109/TNNLS.2016.2616145. Epub 2016 Oct. 26 (www.ncbi.nlm.nih.gov/pubmed/27810837); Jolliffe, I. T., “Principal Component Analysis, Second Edition”, Springer 2002, cda.psych.uiuc.edu/statistical_learning_course/Jolliffe I. Principal Component Analysis (2ed., Springer, 2002)(518s)_MVsa_.pdf, Stone, James V. “Blind source separation using temporal predictability.” Neural computation 13, no. 7 (2001):1559-1574; Barros, Allan Kardec, and Andrzej Cichocki. “Extraction of specific signals with temporal structure.” Neural computation 13, no. 9 (2001):1995-2003; Lee, Soo-Young. “Blind source separation and independent component analysis: A review.” Neural Information Processing-Letters and Reviews 6, no. 1 (2005): 1-57; Hyvärinen, Aapo, and Patrik Hoyer. “Emergence of phase- and shift-invariant features by decomposition of natural images into independent feature subspaces.” Neural computation 12, no. 7 (2000):1705-1720; Wahlund, Björn, Wlodzimierz Klonowski, Pawel Stepien, Robert Stepien, Tatjana von Rosen, and Dietrich von Rosen. “EEG data, fractal dimension and multivariate statistics.” Journal of Computer Science and Engineering 3, no. 1 (2010): 10-14; Yu, Xianchuan, Dan Hu, and Jindong Xu. Blind source separation: theory and applications. John Wiley & Sons, 2013; Panda, Shantipriya, Satchidananda Dehuri, and Sung-Bae Cho. “Machine Learning Approaches for Cognitive State Classification and Brain Activity Prediction: A Survey.” Current Bioinformatics 10, no. 4 (2015): 344-359; Friston, Karl J., Andrew P. Holmes, Keith J. Worsley, J-P. Poline, Chris D. Frith, and Richard S J Frackowiak. “Statistical parametric maps in functional imaging: a general linear approach.” Human brain mapping 2, no. 4 (1994):189-210; Wang, Ya n, Matthew T. Sutherland, Lori L Sanfratello, and Akaysha C. Tang. “Single-trial classification of ERPS using second-order blind identification (SOBI).” In Machine Learning and Cybernetics, 2004. Proceedings of 2004 International Conference on, vol. 7, pp. 4246-4251. IEEE, 2004; Jutten, Christian, and Massoud Babaie-Zadeh. “Source separation: Principles, current advances and applications.” IAR Annu Meet Nancy Fr 110 (2006); Saproo, Sameer, Victor Shih, David C Jangraw, and Paul Sajda. “Neural mechanisms underlying catastrophic failure in human-machine interaction during aerial navigation.” Journal of neural engineering 13, no. 6 (2016): 066005; Valente, Giancarlo. “Separazione cieca di sorgenti in ambienti reali: nuovi algoritmi, applicazionie implementazioni.” (2006); SAPIENZA L A. “Blind Source Separation in real-world environments: new algorithms, applications and implementations Separazione cieca di sorgenti in ambienti reali: nuovi algoritmi, applicazionie.”; Ewald, Arne. “Novel multivariate data analysis techniques to determine functionally connected networks within the brain from EEG or MEG data” (2014); Friston, Karl J. “Basic concepts and overview.” SPMcourse, Short course; Crainiceanu, Ciprian M., Ana-Maria Staicu, Shubankar Ray, and Naresh Punjabi. “Statistical inference on the difference in the means of two correlated functional processes: an application to sleep EEG power spectra.” Johns Hopkins University, Dept. of Biostatistics Working Papers (2011): 225; Konar, Amit, and Aruna Chakraborty. Emotion recognition: A pattern analysis approach. John Wiley & Sons, 2014; Kohl, Florian. “Blind separation of dependent source signals for MEG sensory stimulation experiments.” (2013); Onken, Arno, Jian K Liu, P P Chamanthi R. Karunasekara, loannis Delis, Tim Gollisch, and Stefano Panzeri. “Using matrix and tensor factorizations for the single-trial analysis of population spike trains.” PLoS computational biology 12, no. 11 (2016): e1005189; Tressold, Patrizio, Luciano Pederzoli, Marco Bilucaglia, Patrizio Caini, Pasquale Fedele, Alessandro Ferrini, Simone Melloni, Diana Richeld, Florentina Richeld, and Agostino Accardo. “Brain-to-Brain (Mind-to-Mind) Interaction at Distance: A Confirmatory Study.” (2014). f1000researchdata. s3.amazonaws.com/manuscripts/5914/5adbf847-787a-4fc1-ac04-2e1cd61ca972_4336_-_patrizio tressoldi_v3.pdf?doi=10.12688/f1000research.4336.3; Tsia pa ras, Nikolaos N. “Wavelet analysis in coherence estimation of electroencephalographic signals in children for the detection of dyslexia-related abnormalities.” PhD diss., 2006.

FIG. 9 shows a flowchart of an embodiment of the invention. Asleep stage of a first subject is identified, and then it is determined whether the sleep stage is the desired sleep stage. If not, the first subject is further monitored. If the sleep stage is the one desired, the brain activity of the first subject is captured, reflecting the sleep stage, and the brain activity patterns of the first subject while in the desired sleep stage stored in non-volatile memory 500. The stored brain activity patterns are subsequently retrieved and used to induce the sleep stage in a second subject by replicating the brain activity patterns of the first subject in the second subject by appropriate stimulation of the second subject. The second subject may be monitored to verify that the second subject is in the desired sleep stage.

As shown in FIG. 10, a human brain state or mental state in a subject is modified or altered. In some implementations, a current brainwave pattern of the subject, a phase of a characteristic wave of the current brainwave pattern of the subject, a characteristic timing of a stimulus-response dependent on the mental state, or temporal relationships in monitored neurological or motor patterns of the subject is determined. The desired change in the current brain wave pattern of the subject is determined or defined. A stimulus is applied, e.g., electrical, magnetic, acoustic or ultrasound, sensory, etc., which can be for determining the current state, changing the state, or both. For example, a characteristic timing of a stimulus-response dependent on the mental state may be extracted, or temporal relationships in monitored neurological or motor patterns of the subject determined. The stimulus may be asynchronous, or time-synchronized with respect to the phase state, or dependent on at least the determined temporal relationships. In a closed-loop excitation, the brain wave pattern of the subject after at least one stimulus is monitored or the response parameters, e.g., characteristic timing measured or assessed. The stimulus may be controlled dependent on the observed or monitored changes, indicative of an effective alteration or modification of the brain state or mental state in the subject. FIG. 10 thus shows a flowchart of an embodiment of the invention. A desired mental state is identified 540. The mental state of a subject identified 550, and a phase of a dominant brainwave, characteristic of the mental state of the subject identified 560. A stimulus is applied to the subject to change the mental state of the subject to the desired mental state, while synchronizing the phase of the stimulus with the phase of the dominant brainwave of the subject 570. The subject is monitored to determine if the desired mental state is achieved. If the desired mental state is sleep, the sleep state of the subject may be verified 580.

FIG. 11 shows a flowchart of a further embodiment of the invention. An app is opened on a smartphone, tablet or another mobile or wearable device 1110. Note that in some applications, the device need not be mobile, and for example may be part of a headboard, nightstand, dock radio, etc A soundtrack conducive to sleep, e.g., sounds of rainfall, waterfall, ocean waves, a melody, white noise, pink noise, etc., is chosen 1120. An organic waveform is chosen, derived from brainwaves of a sleeping donor, corresponding to a specific stage of a sleep cycle or a complete sleep cycle 1130. The sound delivery may be chosen to be through a mobile device speaker, earphones, wireless earbuds. If separate sound delivery to each ear, the sound may be isochronic tones or binaural beats 1140, while if not isolated, isochronic tones may be played 1160.

FIG. 12 shows a flowchart of a still further embodiment of the invention. An app may be opened on a smartphone, tablet or wearable device 110. Light settings, such as color and intensity, are chosen 1220. An organic waveform derived from brainwaves of a sleeping donor, corresponding to a specific stage or stages of sleep, or a complete sleep cycle is chosen 1230. Light delivery may be chosen through an ambient light source or e.g., LEDs positioned on a wearable eye mask 1240, which is wirelessly connected to the device. Sleep stimulation is turned on by projecting the light modulated with the chosen organic waveform through ambient light or LEDs positioned near the eyes 1250.

FIG. 13 shows a flowchart of an embodiment of the invention. The subject opens an app on a device 1310, and chooses light and sound settings, e.g., color, intensity, sound, volume, etc. 1320. An organic waveform derived from the brainwaves of a sleeping donor is chosen, e.g., automatically by the app, corresponding to a specificstage(s) of sleep or a complete sleep cycle 1330. The stimulus is chosen as light or sound delivery through the device or wirelessly 1340. Sleep stimulation is turned using synchronized light and sound modulated with the chosen organic waveform 1350.

FIG. 14 shows a flowchart of an embodiment of the invention. The subject opens an app on e.g., a mobile or wearable device 1410, and chooses light and/or sound settings, e.g., color, intensity, sound, volume, etc. 1420. An organic waveform derived from the brainwaves of a sleeping donor is chosen, e.g., automatically by the app, corresponding to a specificstage(s) of sleep or a complete sleep cycle 1430. The stimulus is chosen as light and/or sound delivery through the device or wirelessly 1440. Sleep stimulation is turned using synchronized light and/or sound modulated with the chosen organic waveform 1450. EEG and/or other biometric data is recorded from the subject and transmitted in real time to the device or a cloud computing resource for analysis 1460. The stimulation of the subject is adjusted based on the data received from the subject 1470.

FIG. 15 shows a flowchart of an embodiment of the invention. The subject opens an app on, e.g., a mobile or wearable device and logs in to a personal account 1510. A new waveform, modified from the last waveform used based on biometric sleep data received from the subject during a previous stimulation session 1520. Light and/or sound delivery through the device or through a wireless peripheral is chosen 1540. Sleep stimulation is turned using synchronized light and/or sound modulated with the chosen organic waveform 1550. EEG and/or other biometric data is recorded from the subject and transmitted to a remote for analysis 1560. The received biometric data from the subject is analyzed, to measure the effectiveness of the stimulation and to adjust the waveform accordingly, to improve the effect of the stimulation 1570.

Therefore, statistical approaches are available for separating EEG signals from other signals, and for analyzing components of EEG signals themselves. According to the present invention, various components that might be considered noise in other contexts, e.g., according to prior technologies, such as a modulation pattern of a brainwave, are preserved. Likewise, interactions and characteristic delays between significant brainwave events are preserved. This information may be stored either integrated with the brainwave pattern in which it occurs or as a separated modulation pattern that can then be recombined with an unmodulated brainwave pattern to approximate the original subject.

According to the present technology, lossy “perceptual” encoding (i.e., functionally optimized with respect to a subjective response) of the brainwaves may be employed to process, store, and communicate the brainwave information. In a testing scenario, the “perceptual” features may be tested, so that important information is preserved over information that does not strongly correspond to the effective signal. Thus, while one might not know a priori which components represent useful information, a genetic algorithm may empirically determine which features or data reduction algorithms or parameter sets optimize retention of useful information vs. information efficiency. It is noted that subjects may differ in their response to signal components, and therefore the “perceptual” encoding may be subjective with respect to the recipient. On the other hand, different donors may have different information patterns, and therefore, each donor may also require individual processing. As a result, pairs of donor and recipient may require optimization, to ensure accurate and efficient communication of the relevant information. According to the present invention, sleep/wake mental states and their corresponding patterns are sought to be transferred. In the recipient, these patterns have characteristic brainwave patterns. Thus, the donor may be used, under a variety of alternate processing schemes, to stimulate the recipient, and the sleep/wake response of the recipient determined based on objective criteria, such as resulting brainwave patterns or expert observer reports, or subjective criteria, such as recipient self-reporting, survey or feedback. Thus, after a training period, optimized processing of the donor, which may include filtering, dominant frequency resynthesis, feature extraction, etc., may be employed, which is optimized for both donor and recipient. In other cases, the donor characteristics may be sufficiently normalized, that only recipient characteristics need be compensated. In a trivial case, there is only one exemplar donor, and the signal is oversampled and losslessly recorded, leaving only recipient variation as a significant factor.

Because dominant frequencies tend to have low information content (as compared to the modulation of these frequencies and interrelation of various sources within the brain), one efficient way to encode the main frequencies is by location, frequency, phase, and amplitude. The modulation of a wave may also be represented as a set of parameters. By decomposing the brainwaves according to functional attributes, it becomes possible, during stimulation, to modify the sequence of “events” from the donor, so that the recipient need not experience the same events, in the same order, and in the same duration, as the donor. Rather, a high-level control may select states, dwell times, and transitions between states, based on classified patterns of the donor brainwaves. The extraction and analysis of the brainwaves of the donors, and response of the recipient, may be performed using statistical processes, such as principal components analysis (PCA), independent component analysis (ICA), and related techniques; clustering, classification, dimensionality reduction and related techniques; neural networks and other known technologies. These algorithms may be implemented on general purpose CPUs, array processors such as GPUs, and other technologies.

In practice, a brainwave pattern of the first subject may be analyzed by a PCA technique that respects the non-linearity and non-independence of the brainwave signals, to extract the major cyclic components, their respective modulation patterns, and their respective interrelation. The major cyclic components may be resynthesized by a waveform synthesizer, and thus may be efficiently coded. Further, a waveform synthesizer may modify frequencies or relationships of components from the donor based on normalization and recipient characteristic parameters. For example, the brain of the second subject (recipient) may have characteristic classified brainwave frequencies 3% lower than the donor (or each type of wave may be separately parameterized), and therefore the resynthesis may take this difference into account. The modulation patterns and interrelations may then be reimposed onto the resynthesized patterns. The normalization of the modulation patterns and interrelations may be distinct from the underlying major cyclic components, and this correction may also be made, and the normalized modulation patterns and interrelations included in the resynthesis. If the temporal modifications are not equal, the modulation patterns and interrelations may be decimated or interpolated to provide a correct continuous time sequence of the stimulator. The stimulator may include one or more stimulation channels, which may be implemented as electrical, magnetic, auditory, visual, tactile, or another stimulus, and/or combinations.

The stimulator is preferably feedback controlled. The feedback may relate to the brainwave pattern of the recipient, and/or context or ancillary biometric basis. For example, if the second subject (recipient) begins to awaken from sleep, which differs from the first subject (donor) sleep pattern, then the stimulator may resynchronize based on this finding. That is, the stimulator control will enter a mode corresponding to the actual state of the recipient, and seek to guide the recipient to the desired state from a current state, using the available range and set of stimulation parameters. The feedback may also be used to tune the stimulator, to minimize error from a predicted or desired state of the recipient subject based on the prior and current stimulation.

The control for the stimulator is preferably adaptive and may employ a genetic algorithm to improve performance overtime. For example, if there are multiple first subjects (donors), the second subject (recipient) may be matched with those donors from whose brainwave signals (or algorithmically modified versions thereof) the predicted response in the recipient is best, and distinguished from those donors from whose brainwave signals the predicted response in the recipient subject poorly corresponds. Similarly, if the donors have brainwave patterns determined over a range of time and context and stored in a database, the selection of alternates from the database may be optimized to ensure best correspondence of the recipient subject to the desired response.

It is noted that a resynthesizer-based stimulator is not required, if a signal pattern from a donor is available that properly corresponds to the recipient and permits a sufficiently low error between the desired response and the actual response. For example, if a donor and a recipient are the same subject at different times, a large database may be unnecessary, and the stimulation signal may be a minimally processed recording of the same subject at an earlier time. Likewise, in some cases, a deviation is tolerable, and an exemplar signal may be emitted, with relatively slow periodic correction. For example, a sleep signal may be derived from a single subject and replayed with a periodicity of 90 minutes or 180 minutes, such as a light or sound signal, which may be useful in a dormitory setting, where individual feedback is unavailable or unhelpful.

In some cases, it is useful to provide a stimulator and feedback-based controller on the donor. This will better match the conditions of the donor and recipient, and further allow determination of not only the brainwave pattern of the donor but also responsivity of the donor to the feedback. One difference between the donors and the recipients is that in the donor, the natural sleep pattern is sought to be maintained and not interrupted. Thus, the adaptive mufti-subject database may include data records from all subject, whether selected ab initio as a useful exemplar or not. Therefore, the issue is whether a predictable and useful response can be induced in the recipient from the database record and, if so, that record may be employed. If the record would produce an unpredictable result or a non-useful result, the use of that record should be avoided. The predictability and usefulness of the responses may be determined by a genetic algorithm or other parameter-space searching technology.

Extending the sleep signal illumination example, an illuminator (e.g., red LED lightbulb) may have an intensity modulated based on a donors' brainwave pattern. The illuminator may have a flash memory module with tens or hundreds of different brainwave patterns available. The illuminator may further include a sensor, such as a camera or non-imaging optical or infrared sensor, and speech control, similar to Amazon Alexa. The illuminator may also include an associated speaker, to play synchronized sounds or music. When a sleep cycle is commenced, the illuminator begins displaying (and playing and associated audio) the brainwave pattern as a program, seeking to induce a predetermined sleep pattern. The sensors may be used to determine whether the recipient is in the predicted sleep state based on the program. If the recipient has a sleep state that deviates from the program, then the program may be reset to a portion that corresponds to the actual state of the recipient or reset to a guiding state that seeks to guide the sleep state of the recipient back to the desired program. If the target subject cannot be efficiently synchronized or guided, then the illuminator may adopt a different source subject brainwave pattern. In this case, no electrical stimulation or electrical feedback is employed, and the entire operation may be non-contact.

FIG. 16 shows a representation of a mobile device 11. The mobile device is shown in a familiar “smadphone” form factor. Data can be transferred to and from the mobile device 11 via wireless data communications. In general, the mobile device 11 can include a touch-sensitive display screen 18, a speaker 30, a microphone 31, and one or more control buttons 32 for controlling some operations of device 11. The device 11 depicted in FIG. 1(a) can be a device, such as, for example, a smartphone capable of communicating with a wireless local area network, and so forth. In this respect, the mobile device 11 can be implemented with touch screen capabilities associated with the display screen 18. Display screen 18 can be configured to display data including video and text and icons 33 operable as soft buttons providing options and action by the mobile device 11 when selected by a user. The mobile device 11 can be capable of carrying out a variety of functionalities. For example, microprocessor shown as CPU 10 of the mobile device 11 can function as the main controller operating under the control of operating docks supplied from a dock oscillator. CPU 10 can be configured as, for example, a microprocessor. Such a microprocessor can be configured to facilitate the operations of and communicate by the electronic wireless hand-held multimedia device 11. External pins of CPU 10 can be coupled to an internal bus 26 so that it can be interconnected to respective components. The mobile device 11 can also be configured to include memories such as, for example, SRAM 24 which can be provided as a writeable memory that does not require a refresh operation and can be generally utilized as a working area of CPU 10, SRAM (Static RAM) is generally a form of semiconductor memory (RAM) based on a logic circuit known as a flip-flop, which retains information as long as there is enough power to run the device. Font ROM 22 can be configured as a read only memory for storing character images (e.g., icons and font) displayable on a display 18, which can be implemented as, for example, a touch-sensitive display screen. Example types of displays that can be utilized in accordance with display 18 include, for example, a TFT active matrix display, an illuminated LCD (Liquid Crystal Display), or other small-scaled displays being developed or available in the art in compact form. CPU 10 can be utilized to drive display 18 utilizing, among other media, font images from Font ROM 22 and images transmitted as data through wireless unit 17 and processed by image-processing unit 35. EPROM 20 can be configured as a read-only memory that is generally erasable under certain conditions and can be utilized for permanently storing control codes for operating respective hardware components and security data, such as a serial number. A camera capable of capturing video and pictures can be provided and can also work in conjunction with the image processing unit 35.

IR controller 14, when provided, can be generally configured as a dedicated controller for processing infrared codes transmitted/received by an IR transceiver module 16 and for capturing the same as computer data. Wireless unit 17 can be generally configured as a dedicated controller and transceiver module for processing all wireless data transmitted from and to a wireless communications network. It can be appreciated that other variations for wireless transceiver module 17 can also be provided, such as standardized Bluetooth, NFC, Zigbee, etc., and proprietary RF protocols that may be developed for specialized applications.

Port 12 can be connected to CPU 10 and can be temporarily attached, for example, to a docking station to transmit information to and from the mobile device 11 to other devices, such as personal computers. In light of the present invention, port 12 can also be connected to external probes and external sensors for monitoring or providing data. Port 12 can also be configured, for example to link with a battery charger, data communication device, and can permit network devices, a personal computer, or other computing devices to communicate with mobile device 11.

User controls 32 can permit a user to enter data to mobile device 11 and/or initiate particular processing operations via CPU 10. A user interface 33 can be linked to user controls 32 to permit a user to access and manipulate electronic wireless hand held multimedia device 11 for a particular purpose, such as, for example, viewing video images on display 18. User interface 33 can be implemented as a touch screen manipulated user interface, as indicated by the dashed lines linking display 18 with user interface 33. User interface 33 can be configured to accept user input into the mobile device 11. In addition, CPU 10 can cause a sound generator 28 to generate sounds of predetermined frequencies from a speaker 30. Speaker 30 can be utilized to produce music and other audio information associated with video data transmitted to mobile device 11 from an outside source.

A GPS (Global Positioning System) module 13 can be included in the mobile device and can be connected to bus 26. GPS module 13 can be configured to provide location information for the mobile device 11 and can operate with mapping software and resources to provide navigable directions on the display screen 18 to the user, which can be referred to as GPS mapping. The CPU 10 can execute “apps”, which are downloadable programs that provide a user interface, and access to various application programming interface (API) calls made available through the operating system, but are generally limited to executing in a low privilege mode and without direct hardware or driver level access. The apps may be downloaded from the Internet, or an on-line service (e.g., iTunes store, Google Play) or through a wireless transceiver.

FIG. 17 shows a hypnogram of a healthy adult. As shown, the sleep cycle progresses non-monotonically through a series of stages.

FIG. 18 shows a hypnogram of a healthy adult. As shown, one sleep cycle lasting approximately 90 min is comprised of several sleep stages, including REM sleep (R), first non-REM stage (N1), second non-REM stage (N2), and third non-REM stage (N3), also known as slow-wave sleep, having different duration and periodicity. The waking stage is indicated on the hypnogram as W.

FIG. 19 shows a flowchart indicating the sequence of sleep stages.

FIGS. 20A-20C show a sample of the REM stage of sleep in a 34 year-old female under different filtering. This sample is obtained from the database of Sleep EEG recordings described in B Kemp, A H Zwinderman, B Tuk, H A C Kamphuisen, J J L Oberyé. Analysis of a sleep-dependent neuronal feedback loop: the slow-wave microcontinuity of the EEG. IEEE-BME 47(9):1185-1194 (2000) has been used. For each sleep/wake state of each subjects, 1260-seconds samples have been obtained (totaling up to 72 samples per subject, totaling 4898 samples). Only one channel (Fpz-Cz) has been considered. The samples have been cleaned from noise/non-stationary component using singular spectrum analysis, see Singular Spectrum Analysis with R. Springer. 2018 Authors: Golyandina, Nina, Korobeynikov, Anton, Zhigljaysky, Anatoly, generally following approach of Neurosci Methods. 2016 Nov. 1; 273:96-106. doi:10.1016/j.jneumeth.2016.08.008. Epub 2016 Aug. 12; Improving time-frequency domain sleep EEG classification via singular spectrum analysis. Mahvash Mohammad, Kouchaki, Ghavami, Sanei. Data analysis showed that the use of just 16 55A components is sufficient to preserve waveform spectrum of the EEG recordings, whereas 64 55A components is enough to very precisely match the shape as well. Restoration with the use of 64 components has been used to generate “filtered” EEG samples.

FIG. 20A shows the original, FIG. 20B shows sample reconstructed with 64 SSA groups, and FIG. 20C shows the sample reconstructed with 16 SSA groups, for a sample of REM phase of 34-year female.

FIGS. 21A and 21B show an EEG for a 30 years old female in sleep stage R.

FIG. 22 show an EEG for a 30 years old female in sleep stage 3.

FIGS. 23A and 23B show an EEG for a 30 years old female in sleep stage 3.

FIGS. 24A and 24B show an EEG for a 25 years old female in sleep stage W.

FIGS. 25A and 25B show an EEG for a 25 years old male in sleep stage 2.

FIGS. 26A and 26B show an EEG for a 25 years old male in sleep stage 1.

FIGS. 27A and 27B show an EEG for a 25 years old male in sleep stage W.

See Reference List Table 19

Through the whole document, the term “connected to” or “coupled to” that is used to designate a connection or coupling of one element to another element includes both a case that an element is “directly connected or coupled to” another element and a case that an element is “electronically connected or coupled to” another element via still another element. Further, it is to be understood that the term “comprises or includes” and/or “comprising or including” used in the document means that one or more other components, steps, operation and/or existence or addition of elements are not excluded in addition to the described components, steps, operation and/or elements unless context dictates otherwise.

Through the whole document, the term “unit” or “module” includes a unit implemented by hardware or software and a unit implemented by both of them. One unit may be implemented by two or more pieces of hardware, and two or more units may be implemented by one piece of hardware.

Other devices, apparatus, systems, methods, features, and advantages of the invention will be or will become apparent to one with skill in the art upon examination of the following figures and detailed description. It is intended that all such additional systems, methods, features, and advantages be included within this description, be within the scope of the invention, and be protected by the accompanying claims.

In this description, several preferred embodiments were discussed. Persons skilled in the ad will, undoubtedly, have other ideas as to how the systems and methods described herein may be used. It is understood that this broad invention is not limited to the embodiments discussed herein. Rather, the invention is limited only by the following claims.

The aspects of the invention are intended to be separable and may be implemented in combination, sub-combination, and with various permutations of embodiments. Therefore, the various disclosure herein, including that which is represented by acknowledged prior art, may be combined, sub-combined and permuted in accordance with the teachings hereof, without departing from the spirit and scope of the invention. All references and information sources cited herein are expressly incorporated herein by reference in their entirety.

• • Each reference is expressly incorporated herein by reference in its entirety

Claims

1. A method of facilitating sleep using brain stimulation, comprising:

providing data defining a plurality of waveforms in a memory;

retrieving a selected waveform from the memory, selectively dependent on at least one of a determined sleep phase of a human subject and a predetermined sequence; and

stimulating the human subject with a stimulus modulated according to the selected waveform, to facilitate sleep in the human subject.

2. The method according to claim 1, wherein the plurality of waveforms in the memory are derived from brain activity measurements acquired during at least one sleep cycle of at least one human.

3. The method according to claim 1, wherein the plurality of waveforms in the memory are derived from brain activity measurements acquired during at least one sleep cycle of the human subject.

4. The method according to claim 1, further comprising:

acquiring brain neuronal activity measurements during at least one sleep cycle of at least one human; and

processing the acquired brain activity measurements to define the plurality of waveforms in the memory;

entaining the brain of the human subject with the selected waveform.

5. The method according to claim 1, wherein the stimulus is modulated in a user device associated with the human subject, according to a sleep app stored within the user device, the sleep app being downloadable and upgradeable from a remote server.

6. The method according to claim 5, wherein the predetermined sequence is defined by a human user interface menu of the user device for selecting at least one respective waveform.

7. The method according to claim 1, wherein the sleep phase of the human subject is determined based on at least neuronal activity of the human subject recorded via an electroencephalogram.

8. The method according to claim 1, wherein the sleep phase of the human subject is determined based on at least bioelectric signals received from the human subject.

9. The method according to claim 1, wherein the stimulus modulated according to the selected waveform entrains the brain of the human subject with the selected waveform to facilitate sleep in the human subject.

10. The method according to claim 1, wherein the stimulus modulated according to the selected waveform is at one least of alight stimulus and a sound stimulus.

11. The method according to claim 1, wherein the selected waveform corresponds to at least one specific stage of sleep.

12. The method according to claim 1, wherein each of the plurality of waveforms is derived from recordings of brainwaves of at least one sleeping donor, processed using a statistical decision analysis.

13. The method according to claim 1, further comprising:

adaptively defining a sequence of sleep stages dependent on biometric information received from a sleeping human subject; and

selecting waveforms from the memory in dependence on a correspondence to a respective sleep stage of the adaptively defined sequence of sleep stages;

wherein said stimulating the human subject comprises altering a sleep stage of the human subject dependent on both the determined sleep phase of a human subject and the adaptively defined sequence of sleep stages.

14. The method according to claim 1, wherein the human subject is stimulated with at least one audio transducer and wherein the stimulus comprises at least one of an isochronic tone and binaural beats.

15. The method according to claim 1, wherein the human subject is stimulated with an ambient light stimulus, selectively modulated according to the selected waveform to change at least one of brightness and color.

16. The method according to claim 15, wherein the ambient light stimulus is emitted by at least one light emitting diode disposed in a sleep mask proximate the human subjects eyes.

17. The method according to claim 1, further comprising providing at least one sensor to determine at least one of an eye movement and a facial expression of the human subject to at least one of determine a current determined sleep phase of a human subject or select the predetermined sequence.

18. The method of claim 1, wherein the predetermined sequence is a natural series of sleep stages, the method further comprising resetting the progress according to the natural series of sleep stages in dependence on an awakening of the human subject.

19. A method of generating a waveform for neuromodulation to improve sleep in a subject, the method comprising:

collecting EEG recordings from at least one sleeping donor for a plurality of sleep stages;

grouping a plurality of portions of the EEG recordings corresponding to the plurality of sleep stages, into a plurality of groups corresponding to the plurality of sleep stages;

analyzing each group using a statistical analysis;

extracting a cortical signature corresponding characteristic of each analyzed group;

generating a waveform based on the characteristic cortical signature for each of the plurality of sleep stages; and

modulating a stimulus for the subject according to the generated waveforms for the plurality of sleep stages.

20. A mobile device contained within a housing, comprising:

a microprocessor;

an electrical power source, electrically coupled with the microprocessor;

a wireless communication transceiver, electrically coupled with the microprocessor;

at least one microphone port, electrically coupled with the microprocessor, configured to receive an electrical signal corresponding to a sound;

at least one camera port electrically coupled with the microprocessor, configured to receive an electrical signal corresponding to an image;

a display, electrically coupled with the microprocessor;

at least one speaker port, electrically coupled with the microprocessor, configured to generate an electrical signal corresponding to a sound;

a non-volatile memory and electrically coupled with the microprocessor, configured to store at least one app downloadable through the wireless communication transceiver for controlling the microprocessor, said at least one downloadable app being configured to:

(a) select a waveform from a plurality of waveforms derived from brainwaves of at least one sleeping donor, wherein said waveform corresponds to at least one a specific stage of sleep, a gender, and an age group; and

(b) define a stimulus for stimulation of a subject, selected from the group consisting of at least one of an auditory stimulus generated through the at least one speaker, and a visual stimulus generated through the display, modulated with the selected waveform, and adapted to entrain the brain of the subject with the selected waveform to facilitate sleep in the subject;

wherein at least one of the selection of the waveform and the definition of the stilulus is responsive to the at least one microphone or the at least one camera.